Patent Application
20190172555
A microbiome is an ecological community of commensal, symbiotic, and pathogenic microorganisms that are associated with an organism. The human microbiome comprises more microbial cells than human cells, but characterization of the human microbiome is still in nascent stages due to limitations in sample processing techniques, genetic analysis techniques, and resources for processing large amounts of data. Nonetheless, the microbiome is suspected to play at least a partial role in a number of health/disease-related states (e.g., preparation for childbirth, diabetes, auto-immune disorders, gastrointestinal disorders, rheumatoid disorders, neurological disorders, etc.).
Given the profound implications of the microbiome in affecting a subject's health, efforts related to the characterization of the microbiome, the generation of insights from the characterization, and the generation of therapeutics configured to rectify states of dysbiosis should be pursued. Current methods and systems for analyzing the microbiomes of humans and providing therapeutic measures based on gained insights have, however, left many questions unanswered. In particular, methods for characterizing certain health conditions and therapies probiotic therapies) tailored to specific subjects based upon microbiome compositional or functional diversity features have not been viable due to limitations in current technologies.
As such, there is a need in the field of microbiology for a new and useful method and system for characterizing health conditions in an individualized and population-wide manner. This invention creates such a new and useful method and system.
In a first aspect, the present invention provides a method for identification and classification of occurrence of a microhiome associated with an oral health issue or screening for the presence or absence of a microhiome associated with an oral health issue in an individual and/or determining a course of treatment for an individual human having a microhiome composition associated with a health condition derived from oral health issue, the method comprising:
In embodiments described herein, reference is made to “bacteria” and “bacterial material” (e.g., DNA). Additionally or alternatively, other microorganisms and their material (e.g., DNA) can he detected, classified, and used in the methods and compositions described herein and thus every occurrence of “bacterial” or “ba.cterial material” or equivalents thereof apply equally to other microorganisms, including but not limited to archaea, unicellular eukaryotic organisms, viruses, or the combinations thereof.
In a second aspect, the present invention provides a method of determining a classification of occurrence of a microbiome indicative of, or associated with, an oral health issue or screening for the presence or absence of a microbiome indicative of an oral health issue in an individual and/or determining a course of treatment for an individual human having a microbiome indicative of an oral health issue, the method comprising, providing a sample comprising bacteria (or at least one of the following microorganisms including: bacteria, archaea, unicellular eukaryotic organisms and viruses, or the combinations thereof) from the individual human; determining an amount( )of one or more of the following in the sample: bacteria taxon or gene sequence corresponding to gene functionality as set forth in TABLEs A or B; comparing the determined amount(s) to a disease signature having cut-off or probability values for amounts of the bacteria taxon and/or gene sequence for an individual having a microbiome indicative of an oral health issue or an individual not having a microbiome indicative of an oral health issue or both; and determining a classification of the presence or absence of the microbiome indicative of an oral health issue and/or determining the course of treatment for the individual human having the microbiome indicative of an oral health issue based on the comparing.
In some embodiments, the oral health issue is: (i) dental decay and the bacteria taxa or gene sequences corresponding to gene functionality are selected from those in TABLE A; or (ii) gingivitis and the bacteria taxa or gene sequences corresponding to gene functionality are selected from those in TABLE B. In some embodiments, the determining comprises preparing DNA from the sample and performing nucleotide sequencing of the DNA. In some embodiments, the determining comprises deep sequencing bacterial DNA from the sample to generate sequencing reads, receiving at a computer system the sequencing reads; and mapping, with the computer system, the reads to bacterial genomes to determine whether the reads map to a sequence from the bacterial taxon or gene sequence corresponding to gene functionality from TABLEs A or B; and determining a relative amount of different sequences in the sample that correspond to a sequence from the bacteria taxon or gene sequence corresponding to gene functionality from TABLEs A or B.
In some embodiments, the deep sequencing is random deep sequencing. In some embodiments, the deep sequencing comprises deep sequencing of 16S rRNA (e.g., bacterial and/or archaeal) coding sequences. In some embodiments, the method further comprises obtaining physiological, demographic or behavioral information from the individual human, wherein the disease signature comprises physiological, demographic or behavioral information; and the determining comprises comparing the obtained physiological, demographic or behavioral information to corresponding information in the disease signature. In some embodiments, the sample is an oral sample from the individual human. In some embodiments, the method further comprises determining that the individual human likely has a microbiome indicative of an oral health issue; and treating the individual human to ameliorate at least one symptom of the microbiome indicative of the oral health issue. In some embodiments, the treating comprises administering a dose of one of more of the bacteria taxon listed in TABLEs A or B to the individual human for which the individual human is deficient.
In a third aspect, the present invention provides a method for determining a classification of the presence or absence of a microbiome indicative of an oral health issue and/or determine a course of treatment for an individual human having a microbiome indicative of an oral health issue, the method comprising performing, by a computer system: receiving sequence reads of bacterial DNA obtained from analyzing a test sample from the individual human; mapping the sequence reads to a bacterial sequence database to obtain a plurality of mapped sequence reads, the bacterial sequence database including a plurality of reference sequences of a plurality of bacteria; assigning the mapped sequence reads to sequence groups based on the mapping to obtain assigned sequence reads assigned to at least one sequence group, wherein a sequence group includes one or more of the plurality of reference sequences; determining a total number of assigned sequence reads; for each sequence group of a disease signature set of one or more sequence groups selected from TABLEs A or B: determining a relative abundance value of assigned sequence reads assigned to the sequence group relative to the total number of assigned sequence reads, the relative abundance values forming a test feature vector; comparing the test feature vector to calibration feature vectors generated from relative abundance values of calibration samples having a known status of oral health; and determining the classification of the presence or absence of the microbiome indicative of an oral health issue and/or determining the course of treatment for the individual human having the microbiome indicative of an oral health issue based on the comparing.
In some embodiments, the comparing includes: clustering the calibration feature vectors into a control cluster not having the microbiome indicative of an oral health issue and a disease cluster having the microbiome indicative of an oral health issue; and determining which cluster the test feature vector belongs. In some embodiments, the clustering includes using a Bray-Curtis dissimilarity, In some embodiments, the comparing includes comparing each of the relative abundance values of the test feature vector to a respective cutoff value determined from the calibration feature vectors generated from the calibration samples. In some embodiments, the comparing includes: comparing a first relative abundance value of the test feature vector to a disease probability distribution to obtain a disease probability for the individual human having a microbiome indicative of an oral health issue, the disease probability distribution determined from a plurality of samples having the microbiome indicative of the oral health issue and exhibiting the sequence group; comparing the first relative abundance value to a control probability distribution to obtain a control probability for the individual human not having a microbiome indicative of an oral health issue, wherein the disease probabilities and the control probabilities are used to determine the classification of the presence or absence of the microbiome indicative of an oral health issue and/or determining the course of treatment for the individual human having the microbiome indicative of an oral health issue.
In some embodiments, the sequence reads are mapped to one or more predetermined regions of the reference sequences. In some embodiments, the disease signature set includes at least one taxonomic group and at least one functional group. In some embodiments, the oral health issue is: (i) dental decay and the sequence groups are selected from those in TABLE A; or (ii) gingivitis and the sequence groups are selected from those in TABLE B. In some embodiments, the analyzing comprises deep sequencing. In some embodiments, the deep sequencing reads are random deep sequencing reads. In some embodiments, the deep sequencing reads comprise 16S rRNA (e.g., bacterial and/or archaeal) deep sequencing reads. In some embodiments, the method further comprises receiving physiological, demographic or behavioral information from the individual human; and using the physiological, demographic or behavioral information in combination with the classification with the comparing of the test feature vector to the calibration feature vectors to determine the classification of the presence or absence of the microbiome indicative of an oral health issue and/or determining the course of treatment for the individual human having the microbiome indicative of an oral health issue. In some embodiments, the method further comprises preparing DNA from the sample and performing nucleotide sequencing of the DNA.
In a fourth aspect, the present invention provides a non-transitory computer readable medium storing a plurality of instructions that when executed, by the computer system, perform any one of the foregoing methods.
In a fifth aspect, the present invention provides a method for at least one of characterizing, diagnosing, and treating an oral health issue in at least a subject, the method comprising: at a sample handling network, receiving an aggregate set of samples from a population of subjects; at a computing system in communication with the sample handling network, generating a microbiome composition dataset and a microbiome functional diversity dataset for the population of subjects upon processing nucleic acid content of each of the aggregate set of samples with a fragmentation operation, a multiplexed amplification operation using a set of primers, a sequencing analysis operation, and an alignment operation; at the computing system, receiving a supplementary dataset, associated with at least a subset of the population of subjects, wherein the supplementary dataset is informative of characteristics associated with the oral health issue; at the computing system, transforming the supplementary dataset and features extracted from at least one of the microbiome composition dataset and the microbiome functional diversity dataset into a characterization model of the oral health issue; based upon the characterization model, generating a therapy model configured to correct the oral health issue; and at an output device associated with the subject and in communication with the computing system, promoting a therapy to the subject with the oral health issue, upon processing a sample from the subject with the characterization model, in accordance with the therapy model.
In some embodiments, generating the characterization model comprises performing a statistical analysis to assess a set of microbiome composition features and microbiome functional features having variations across a first subset of the population of subjects exhibiting the oral health issue and a second subset of the population of subjects not exhibiting the oral health issue. In some embodiments, generating the characterization model comprises: extracting candidate features associated with a set of functional aspects of microbiome components indicated in the microbiome composition dataset to generate the microbiome functional diversity dataset; and characterizing the mental health issue in association with a subset of the set of functional aspects, the subset derived from at least one of clusters of orthologous groups of proteins features, genomic functional features from the Kyoto Encyclopedia of Genes and Genomes (KEGG), chemical functional features, and systemic functional features.
In some embodiments, generating the characterization model of the oral health issue comprises generating a characterization that is diagnostic of at least one symptom of dental decay or gingivitis. In some embodiments, the generating the characterization model of the oral health issue comprises generating a characterization that is diagnostic of at least one symptom of dental decay, and generating a characterization that is diagnostic of at least one symptom of dental decay comprises generating the characterization upon processing the aggregate set of samples and determining presence of features derived from a set of one or more taxa of TABLE A. In sonic embodiments, the generating the characterization model of the oral health issue comprises generating a characterization that is diagnostic of at least one symptom of gingivitis, and generating a characterization that is diagnostic of at least one symptom of gingivitis comprises generating the characterization upon processing the aggregate set of samples and determining presence of features derived from 1) a set of taxa of TABLE B, and 2) a set of one or more functional groups of TABLE B.
In a sixth aspect, the present invention provides a method for characterizing an oral health issue, the method comprising: upon processing an aggregate set of samples from a population of subjects, generating at least one of a microbiome composition dataset and a microbiome functional diversity dataset for the population of subjects, the microbiome functional diversity dataset indicative of systemic functions present in the microbiome components of the aggregate set of samples; at the computing system, transforming at least one of the microbiome composition dataset and the microbiome functional diversity dataset into a characterization model of the oral health issue, wherein the characterization model is diagnostic of the oral health issue producing observed changes in dental and/or gingival health; and based upon the characterization model, generating a therapy model configured to improve a state of the oral health issue.
In some embodiments, generating the characterization comprises analyzing a set of features from the microbiome composition dataset with a statistical analysis, wherein the set of features includes features associated with: relative abundance of different taxonomic groups represented in the microbiome composition dataset, interactions between different taxonomic groups represented in the microbiome composition dataset, and phylogenetic distance between taxonomic groups represented in the microbiome composition dataset. In some embodiments, generating the characterization comprises performing a statistical analysis with at least one of a Kolmogorov-Smirnov test and a t-test to assess a set of microbiome composition features and microbiome functional features having varying degrees of abundance in a first subset of the population of subjects exhibiting the oral health issue and a second subset of the population of subjects not exhibiting the oral health issue, wherein generating the characterization further includes clustering using a Bray-Curtis dissimilarity.
In some embodiments, generating the characterization model comprises generating a characterization that is diagnostic of at least one symptom of a dental decay issue, upon processing the aggregate set of samples and determining presence of features derived from a set of one or more taxa of TABLE A. In some embodiments, generating the characterization model comprises generating a characterization that is diagnostic of at least one symptom of a gingivitis issue, upon processing the aggregate set of samples and determining presence of features derived from 1) a set of taxa of TABLE B, and 2) a set of one or more functional groups of TABLE B. In some embodiments, the method further includes diagnosing a subject with the oral health issue upon processing a sample from the subject with the characterization model; and at an output device associated with the subject, promoting a therapy to the subject with the oral health issue based upon the characterization model and the therapy model.
In some embodiments, promoting the therapy comprises promoting a bacteriophage-based therapy to the subject, the bacteriophage-based therapy providing a bacteriophage component that selectively downregulates a population size of an undesired taxon associated with the oral health issue. In some embodiments, promoting the therapy comprises promoting a prebiotic therapy to the subject, the prebiotic therapy affecting a microorganism component that selectively supports a population size increase of a desired taxon associated with correction of the oral health issue, based on the therapy model. In some embodiments, promoting the therapy comprises promoting a probiotic therapy to the subject, the probiotic therapy affecting a microorganism component of the subject, in promoting correction of the oral health issue, based on the therapy model. In some embodiments, promoting the therapy comprises promoting a microbiome modifying therapy to the subject in order to improve a state of the oral health associated symptom.
The inventors have discovered that characterization of the microbiome of individuals is useful for detecting a microbiome indicative of dental decay or gingivitis. For example, an individual having symptoms indicative of dental decay or gingivitis, or in whom dental decay or gingivitis is suspected, can be tested to confirm or provide further evidence to support or refute a diagnosis of the subject. As another example, an individual can be assayed to determine whether they have a microbiome that is likely to increase the risk of dental decay or gingivitis. As another example, an individual having, or suspected of having, or having a history of, dental decay or gingivitis can be assayed to determine whether the microbiome is likely to be a causative agent, or contribute to the frequency or severity of the dental decay or gingivitis. An individual having symptoms of dental decay or gingivitis, or has dental decay or gingivitis, or has a microbiome (e.g., an oral, gut, or stool microbiome) that causes or contributes to the frequency or severity of dental decay or gingivitis is referred to herein as having an “oral health issue.” Similarly, an individual having symptoms of dental decay, or has dental decay, or has a microbiome (e.g., an oral, gut, or stool microbiome) that causes or contributes to the frequency or severity of dental decay is referred to herein as having a “dental decay issue.” likewise, an individual having symptoms of gingivitis, or has gingivitis, or has a microbiome (e.g., an oral, gut, or stool microbiome) that causes or contributes to the frequency or severity of gingivitis is referred to herein as having a “gingivitis issue.”
Such characterizations are also useful for screening individuals for and/or determining a course of treatment for an individual that has an oral health issue. For example, by deep sequencing bacterial DNAs from control (healthy, or at least not having an oral health issue) individuals and diseased individuals (having an oral health issue), the inventors have discovered that the amount of certain bacteria and/or bacterial sequences corresponding to certain genetic pathways can be used to predict the presence or absence of an oral health issue. The bacteria and genetic pathways in some cases are present in a certain abundance in individuals having an oral health issue, or having a specific oral health issue, as discussed in more detail below whereas the bacteria and genetic pathways are at a statistically different abundance in control individuals that do not have an oral health issue, or do not have a specific oral health issue,
Details of these associations for the specific oral health issue dental decay can be found in TABLE A for bacteria groups (also called taxonomic groups) and or genetic pathways (also called functional groups). Collectively, the taxonomic groups and functional groups are referred to as features, or as sequence groups in the context of determining an amount of sequence reads corresponding to a. particular group (feature). Scoring of a particular bacteria or genetic pathway can be determined according to a comparison of an abundance value to one or more reference (calibration) abundance values for known samples, e.g., where a detected abundance value less than a certain value is associated with a dental decay issue and above the certain value is scored as associated with a lack of a dental decay issue, depending on the particular criterion. Similarly, depending on the particular criterion, a detected abundance value greater than a certain value can be associated with a dental decay issue and below the certain value can be scored as associated with a lack of a dental decay issue or a microbiome that is not indicative of a dental decay issue. The scoring for various bacteria or genetic pathways can be combined to provide a classification for a subject.
Details of these associations for the specific oral health issue gingivitis can be found in TABLE B for bacteria groups (also called taxonomic groups) and or genetic pathways (also called functional groups). Scoring of a particular bacteria or genetic pathway can be determined according to a comparison of an abundance value to one or more reference (calibration) abundance values for known samples, e.g., where a detected abundance value less than a certain value is associated with a gingivitis issue and above the certain value is scored as associated with a lack of a gingivitis issue, depending on the particular criterion. Similarly, depending on the particular criterion, a detected abundance value greater than a certain value can be associated with a gingivitis issue and below the certain value can be scored as associated with a lack of a gingivitis issue or a microbiome that is not indicative of a gingivitis issue. The scoring for various bacteria or genetic pathways can be combined to provide a classification for a subject.
Cardiobacterium
hominis_2718
The comparison of an abundance value to one or more reference abundance values can involve a comparison to a cutoff value determined from the one or more reference values. Such cutoff value(s) can be part of a decision tree or a clustering technique (where a cutoff value is used to determine which cluster the abundance value(s) belong) that are determined using the reference abundance values. The comparison can include intermediate determination of other values, e.g., probability values. The comparison can also include a comparison of an abundance value to a probability distribution of the reference abundance values, and thus a comparison to probability values.
The inventors have identified the specific bacteria taxa and genetic pathways listed in TABLE A by deep sequencing of bacterial DNA associated with samples from test individuals having a dental decay issue and control individuals that do not have a dental decay issue and determining those criteria that readily distinguish test individuals from control individuals. Similarly, the inventors have identified the specific bacteria taxa and genetic pathways listed in
TABLE B by deep sequencing of bacterial DNA associated with samples from test individuals having a gingivitis issue and control individuals that do riot have a gingivitis issue and determining -those criteria that readily distinguish test individuals from control individuals.
Deep sequencing allows for determination of a sufficient number of copies of DNA sequences to determine relative amount of corresponding bacteria or genetic pathways in the sample. flaying identified the criteria in TABLEs A and B, one can now detect an individual that has an oral health issue by detecting one or more (e.g., 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, or more) of the options in TABLE A or B by any quantitative detection method. In some cases, one can now detect an individual that has an oral health issue by detecting from about Ito about 20, from about 2 to about 15, from about 3 to about 10, from about 1 to about 10, from about 1 to about 15, from about 1 to about 5, or from about 5 to about 30 of the options in TABLE A or TABLE B by any quantitative detection method. For example, while deep sequencing can be used to detect the presence, absence or amount of one or more option in TABLE A or TABLE B, one can also use other detection methods, including but not limited to protein detection methods. For example, without intending to limit the scope of the invention, one could use protein-based diagnostics such as immunoassays to detect bacterial taxons by detecting taxon-specific protein markers.
As a result of these discoveries (e.g., as set forth in TABLEs A and B), one can design treatments to ameliorate one or more symptoms of an oral health issue and/or alleviate or reduce the frequency and/or severity of dental decay or gingivitis. As a non-limiting example, one can determine whether an individual having a dental decay issue lacks, or has a reduced abundance of, one or more type of bacteria as listed in TABLE A and if so, that one or more type of bacteria can be administered to the individual. Additionally, or alternatively, one can determine whether an individual having a dental decay issue lacks, or has a reduced abundance of, one or more type of bacteria as listed in TABLE A and if so, a prebiotic that promotes the growth of that one or more type of bacteria can be administered to the individual. Additionally, or alternatively, one can determine whether an individual having a dental decay issue has an increased abundance of one or more type of bacteria as listed in TABLE A and if so, a targeted therapy that reduces the abundance of such bacteria (e.g., bacteriophage therapy or selective antibiotic therapy) can be administered to the individual.
As another non-limiting example, one can determine whether an individual having a gingivitis issue lacks, or has a reduced abundance of, one or more type of bacteria as listed in TABLE B and if so, that one or more type of bacteria can be administered to the individual. Additionally, or alternatively, one can determine whether an individual having a gingivitis issue lacks, or has a reduced abundance of, one or more type of bacteria as listed in TABLE B and if so, a prebiotic that promotes the growth of that one or more type of bacteria can be administered to the individual. Additionally, or alternatively, one can determine whether an individual having a gingivitis issue has an increased abundance of one or more type of bacteria as listed in TABLE B and if so, a targeted therapy that reduces the abundance of such bacteria (e.g., bacteriophage therapy or selective antibiotic therapy) can be administered to the individual.
In some embodiments, a method of determining whether, or the likelihood whether, an individual has an oral health issue is provided. As described herein, an individual having an oral health issue can exhibit an increase in one or more taxonomic groups in the microbiome, a decrease in one or more taxonomic groups in the microbiome, an increase in one or more functional groups in the microbiome, a decrease in one or more functional groups in the microbiome, or a combination thereof (e.g., relative to a control/healthy individual or population of control or healthy individuals).
The method can include one or more of the following steps:
The likelihood of an individual having a microbiotne indicative of an oral health issue (e.g., as listed in TABLEs A or B) refers to the chance (degree of confidence) that the results from the individual's sample can be correlated with an oral health issue. Alternatively, one can simply screen for an oral health issue, i.e., one can generate a yes or no indication for the presence or absence of a microbiome indicative of dental decay or gingivitis. In some embodiments, the individual will not yet have been diagnosed with dental decay or gingivitis or a dental decay issue or gingivitis issue. In other examples, the individual can have been initially diagnosed by other methods and the methods described herein can be used to provide better (or worse) confidence of the initial diagnosis.
Any type of sample containing bacteria can be used from the individual. Exemplary sample types include, for example, a fecal sample, blood sample, saliva sample, throat swab, cheek swab, gum swab, urine or other bodily fluid from the individual. Nucleic acids (e.g., DNA and/or RNA) can be purified from the sample. Basic texts disclosing the general molecular biology methods include Sambrook and Russell, Molecular Cloning, A Laboratory Manual (3rd ed. 2001); Kriegler, Gene Transfer and Expression: A Laboratory Manual (1990); and Current Protocols in Molecular Biology (Ausubel et al., eds., 1994-1999). Such nucleic acids may also be obtained through in vitro amplification methods such as those described herein and in Berger, Sambrook, and Ausubel, as well as Mullis et ch., (1987) U.S. Pat. No. 4,683,202; PCR Protocols A Guide to Methods and Applications (Innis et al., eds) Academic Press Inc. San Diego, Calif. (1990) (Innis); Arnheim Levinson (Oct. 1, 1990) C&EN 36-47; The Journal Of NIH Research (1991) 3: 81-94; Kwoh et al, (1989) Proc. Nail. Acad. Sci. USA 86: 1173; Guatelli et al. (1990) Proc. Natl. Acad. Sci. USA 87, 1874; Lomell et al. (1989) J. Clin. Chem., 35: 1826; Landegren et al., (1988) Science 241: 1077-1080; Van Brunt (1990) Biotechnology 8: 291-294; Wu and Wallace (1989) Gene 4: 560; and Barringer et al. (1990) Gene 89: 117, each of which is incorporated by reference in its entirety for all purposes and in particular for all teachings related to amplification methods. In some embodiments, the nucleic acids will not be amplified before they are quantified.
Any of a variety of detection methods can be used to screen an individual's sample for one or more of the features listed in TABLES A or B. For example, in some embodiments, nucleic acid hybridization and/or amplification methods are used to detect and quantify one or more of the features. In some embodiments, an immunoassay or other assay to detect and quantify one or more specific proteins determinative of one or more of the criteria can be used.
For example, solid-phase ELISA immunoassays, Western blots, or immunohistochemistry are routinely used to specifically detect a protein. See, Harlow and Lane Antibodies, A Laboratory Manual, Cold Spring Harbor Publications, NY (1988) for a description of immunoassay formats and conditions that can be used to determine specific immunoreactivity. In some preferred embodiments, nucleotide sequencing is used to identify and quantify one or more of the criteria.
DNA sequencing can be performed as desired. Such sequencing can be performed using known sequencing methodologies, e.g., Illumina, Life TechnologiesLife Technologies, and Roche 454 sequencing systems. In typical embodiments, a sample is sequenced using a large-scale sequencing method that provides the ability to obtain sequence information from many reads. Such sequencing platforms include those commercialized by Roche 454 Life Sciences (GS systems), Illumina (e.g., HiSeq, MiSeq) and Life Technologies (e.g., SOLiD systems).
The Roche 454 Life Sciences sequencing platform involves using emulsion PCR and immobilizing DNA fragments onto bead. Incorporation of nucleotides during synthesis is detected by measuring light that is generated when a nucleotide is incorporated.
The Illumina technology involves the attachment of genomic DNA to a planar, optically transparent surface. Attached DNA fragments are extended and bridge amplified to create an ultra-high density sequencing flow cell with clusters containing copies of the same template. These templates are sequenced using a sequencing-by-synthesis technology that employs reversible terminators with removable fluorescent dyes.
Methods that employ sequencing by hybridization may also be used. Such methods, e.g., used in the Life Technologies SOLiD4+ technology uses a pool of all possible oligonucleotides of a fixed length, labeled according to the sequence. Oligonucleotides are annealed and ligated; the preferential ligation by DNA ligase for matching sequences results in a signal informative of the nucleotide at that position.
The sequence can be determined using any other DNA sequencing method including, e.g., methods that use semiconductor technology to detect nucleotides that are incorporated into an extended primer by measuring changes in current that occur when a nucleotide is incorporated (see, e.g., U.S. Patent Application Publication Nos. 20090127589 and 20100035252). Other techniques include direct label-free exonuclease sequencing in which nucleotides cleaved from the nucleic acid are detected by passing through a nanopore (Oxford Nanopore) (Clark et al., Nature nanotechnology 4: 265-270, 2009), and Single Molecule Real time (SMRT™) DNA sequencing technology (Pacific Biosciences), which is a sequencing-by synthesis technique.
Deep sequencing can be used to quantify the number of copies of a particular sequence in a sample and then also be used to determine the relative abundance of different sequences in a sample. Deep sequencing refers to highly redundant sequencing of a nucleic acid sequence, for example such that the original number of copies of a sequence in a sample can be determined or estimated. The redundancy (i.e., depth) of the sequencing is determined by the length of the sequence to be determined (X), the number of sequencing reads (N), and the average read length (L). The redundancy is then NxL/X. The sequencing depth can be, or be at least about 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55 , 56, 57, 58, 59, 60, 70, 80, 90, 100, 110, 120, 130, 150, 200, 300, 500, 500, 700, 1000, 2000, 3000, 4000, 5000 or more. See, e.g, Mirebrahim, Hamid et a, Bioinformatics 31 (12): i9-i16 (2015).
In some embodiments, specific sequences in the sample can be targeted for amplification and/or sequencing. For example, specific primers can be used to detect and sequence bacterial sequences of interest. Exemplary target sequences can include, but are not limited to, the 16S rRNA coding sequence (e.g., gene families mentioned in the discussion of Block S120), as well as gene sequences involved in one or more genetic pathway as shown in TABLE B. In addition, or alternatively, whole genome sequencing methods that randomly sequence DNA fragments in a sample can be used.
Once sequencing raw data is generated, the resulting sequence reads can be “mapped” to known sequences in a genomic database. Exemplary algorithms that are suitable for determining percent sequence identity and sequence similarity and thus aligning and identifying sequence reads are the BLAST and BLAST 2.0 algorithms, which are described in Altschul et al, (1990) j. Mol. Biol. 215: 403-410 and Altschul et al. (1977) Nucleic Acids Res. 25: 3389-3402, respectively. Software for performing BLAST analyses is publicly available through the National Center for Biotechnology Information (NCBI) web site. Accordingly, for the sequence reads generated, a subset of these reads will be aligned to one or more bacterial genomes of the bacterial taxa in TABLEs A or B or can be aligned to a gene sequence in any genome that has a genetic function as set forth in TABLE B. For example, one can align a read with a database of bacterial sequences and the read can be designated as from a particular bacteria if that read has the best alignment to a DNA sequence from that bacteria in the database.
Similarly, one can align a read with a database of bacterial sequences and the read can be designated as from a genetic pathway if that read has the best alignment to a DNA sequence from that genetic pathway in the database. For example, one can assign the read to a sequence from a particular Kyoto Encyclopedia of Genes and Genomes (KEGG) category or Clusters of Orthologous Groups (COG) categories. KEGGs are described more at genome.jp/kegg/. COGs are described in, e.g.. Tatusov, et al., Nucleic Acids Res. 2000 Jan 1; 28(1): 33-36. The TABLE provided herein lists various KEGG and COG categories that are correlated with the presence or absence of a microbiome indicative of an oral health issue. Different levels of KEGG or COG categories are provided in TABLE B. Values in TABLEs A and B for particular criteria are proportional values compared to totals at that taxonomic or functional designation level.
Assuming sequencing has occurred at a sufficient depth, one can quantify the number of reads for sequences indicative of the presence of a feature of TABLE A, thereby allowing one to set a value for an estimated amount of one of the criterion. The number of reads or other measures of amount of one of the features can be provided as an absolute or relative value. An example of an absolute value is the number of reads of 16S rRNA coding sequence reads that map to the genus of Bacteroides. Alternatively, relative amounts can be determined. An exemplary relative amount calculation is to determine the amount of 16S rRNA coding sequence reads for a particular bacterial taxon (e.g., genus , family, order, class, or phylum) relative to the total number of 16S rRNA coding sequence reads assigned to the bacterial domain. A value indicative of amount of a feature in the sample can then be compared to a cut-off value or a probability distribution in a disease signature for a microbiome indicative of an oral health issue. For example, if the signature indicates that a relative amount of feature #1 of 50% or more of all features possible at that level indicates the likelihood of a microbiome indicative of an oral health issue, then quantification of gene sequences associated with feature #1 less than 50% in a sample would indicate a higher likelihood of a microbiome that is not indicative of an oral health issue and alternatively, quantification of gene sequences associated with feature #1 more than 50% in a sample would indicate a higher likelihood of a microbiome indicative of an oral health issue.
Once amounts of various features from TABLEs A or B have been determined and compared to a cut-off or probability value for the corresponding criteria in a disease signature for an oral health issue, one can determine the likelihood of a microbiome indicative of an oral health issue in the individual.
Disease signatures can include criteria corresponding to one or at least one of the features set forth in TABLEs A or B. In some embodiments, 2, 3, or 4 of the criteria of TABLE A can be used in a disease signature for a microbiome indicative of a dental decay issue. In some embodiments, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20 or more (e.g., all) of the criteria of TABLE B can be used in a disease signature for a microbiome indicative of a gingivitis issue.
In some embodiments, supplementary information about the individual can also be used in the disease signature and thus also for determining the likelihood of occurrence of a microbiome indicative of an oral health issue in the individual. Supplementary information can include, for example, different demographics (e.g., genders, ages, marital statuses, ethnicities, nationalities, socioeconomic statuses, sexual orientations, etc.), different health conditions (e.g., health and disease states), different living situations (e.g., living alone, living with pets, living with a significant other, living with children, etc.), different dietary habits (e.g., omnivorous, vegetarian, vegan, sugar consumption, acid consumption, etc.), different behavioral tendencies (e.g., levels of physical activity, drug use, alcohol use, etc.), different levels of mobility (e.g., related to distance traveled within a given time period), biomarker states (e.g., cholesterol levels, lipid levels, etc.), weight, height, body mass index, genotypic factors, and any other suitable trait that has an effect on microbiome composition.
At block 10, a sample comprising bacteria from the individual human is provided. In specific examples, samples can comprise blood samples, saliva samples, plasma/serum samples (e.g., to enable extraction of cell-free DNA), cerebrospinal fluid, and tissue samples.
In some cases, the sample is an oral sample (e.g., a throat, tongue, or gum swab, or saliva), or a sample (e.g., a nucleic acid sample, such as a DNA sample) extracted from an oral sample.
At block 11, an amount(s) of bacteria taxon and/or gene sequence corresponding to gene functionality as set forth in TABLES A or B is determined. As various examples, an amount of one bacteria taxon can be determined; an amount of one gene sequence corresponding to gene functionality can be determined; an amount of one bacteria taxon and an amount one gene sequence corresponding to gene functionality can be determined; multiple amounts (e.g., 2-4) of bacteria taxa can be determined; multiple amounts (e.g., 2-6) of gene sequences corresponding to gene functionalities can be determined; and multiple amounts of both can be determined.
The amount can be determined in various ways, e.g., by sequencing nucleic acids in the sample, using a hybridization array, and PCR. As examples, the amounts can correspond to levels of a signal or a count of numbers of nucleic acids corresponding to each taxa. The amount can be a relative abundance value.
At block 12, the determined amount(s) are compared to a condition signature having cut-off or probability values for amounts of the bacteria taxon and/or gene sequence for an individual having a microbiome indicative of an oral health issue or an individual not having a microbiome indicative of an oral health issue or both. In various embodiments, each amount can be compared to a separate value, and a number of taxa exceeding that value can be compared to a threshold for determining whether a sufficient number of the taxa provide the condition signature. Other examples are provider herein. Before a comparison to a probability value, the amount can be transformed (e.g., via a probability distribution). As another example, the amounts can be used to determine a measure probability, which can be compared to the probability value, which discriminates among classifications.
At block 13, a classification of the presence or absence of the microbiome indicative of an oral health issue is determined based on the comparing, and/or the course of treatment for the individual human having the microbiome indicative of an oral health issue is determined based on the comparing. As described herein, the classification can be binary or includes more levels, e.g., corresponding to a probability.
Also provided are methods of determining a course of treatment, and/or optionally of treating, an individual having a microbiome indicative of an oral health issue. For example, by detecting the presence, absence, or quantity of one or more of the criteria set forth in TABLEs A or B, one can determine treatments to increase those criteria that are reduced in individuals having a condition/disease (i.e., individuals having a microbiome indicative of an oral health issue) or decrease these criteria that are increased in individuals having the disease (an oral health issue) compared to healthy individuals (i.e., individuals having a microbiorne that is not indicative of an oral health issue). In some embodiments, the individual will have been diagnosed, optionally by other methods, of having a microbiome associated with an oral health issue, or symptoms thereof, and the methods described herein (e.g., comparison to the disease signature) will reveal excessive amounts and/or deficient amounts of one or more of the features that can then be used to guide treatment.
For example, in embodiments in which the amount of a particular bacteria type is lower in individuals having a microbiome indicative of an oral health issue than in individuals having a microbiome that is not indicative of an oral health issue, a possible treatment is providing a probiotic or prebiotic treatment that provides or stimulates growth of the particular bacteria type.
In embodiments in which the higher amount of bacteria is in the individual having a microbiome indicative of an oral health issue, one can administer treatments that reduce the relative amount of that particular bacteria. In some embodiments, antibiotics can be administered to reduce the target bacterial population. Alternatively, other treatments can be administered including promoting (by administration of probiotics or prebiotics) bacteria that compete with the target bacteria. In yet another embodiment, bacteriophage targeting the particular bacteria can be administered to the individual.
Similarly, where a particular function (e.g., KEGG or COG category) is indicated, one can increase or reduce that function by selectively promoting or reducing growth of bacterial populations that have that particular function.
Additional mechanisms of treatment are listed, for example, in
Further, one can monitor treatment of an individual having a microbiome indicative of an oral health issue by obtaining samples from the individual before, during, and/or after treatment of the oral health issue, or before, during, and/or after treatment to mitigate the symptoms of an oral health issue (e.g., prebiotic, probiotic, or bacteriophage therapy), or the combination thereof, to monitor progression of the oral health issue (e.g., monitor progression of dental decay or gingivitis) For example, in sonic embodiments, levels of one or more of the criteria in TABLEs A or B are determined one or more (e.g., 2 or more, 3, 4, 5 or more) times and the dosage of a prebiotic and/or probiotic treatment can he adjusted up or down depending on how the criteria respond to the treatment.
In some embodiments, sequence information can he received. The sequence information can correspond to one or more sequence reads per nucleic acid molecule (e.g., a DNA fragment). The sequence reads can be obtained in a variety of ways. For example, a hybridization array, PCR, or sequencing techniques can he used.
When sequencing is performed, a sequence read can be aligned (mapped) to a plurality of reference bacterial genomes (also called reference genomes) to determine which reference bacterial genome the sequence read aligns and where on that reference genome the sequence read aligns. The alignment can be to a particular region (e.g., 16S region) of a reference genome, and thus to a reference sequence, which can be all or part of the reference genome. For paired-end sequencing, both sequence reads can be aligned as a pair, with an expected length of the nucleic acid molecule being used to aid in the alignment.
Accordingly, it can be determined that a particular DNA fragment is derived from a particular gene of a particular bacterial taxonomic group (also called taxon) based on the aligned location of a sequence read to the particular gene of the particular bacterial taxonomic group. The same determination may be made by various hybridization probes using a variety of techniques, as will be known by one skilled in the art. Thus, the mapping can be performed in a variety of ways.
In this manner, a count of the number of sequence reads aligned to each of one or more genes of different bacterial taxonomic groups can be determined. The count for each gene and for each taxonomic group can be used to determine relative abundances. For example, a relative abundance value (RAV) of a particular taxonomic group can be determined based on a fraction (proportion) of sequence reads aligning to that ta.xonomic group relative to other taxonomic groups. The RAV can correspond to the proportion of reads assigned to a particular ta.xonomic or functional group. The proportion can be relative to various denominator values, e.g., relative to all of the sequence reads, relative to all assigned to at least one group (taxonomic or functional), or all assigned to for a given level in the hierarchy. The alignment can he implemented in any manner that can assign a sequence read to a particular taxonomic or functional group. For example, based on the mappings to the reference sequence(s) in the 16S region, a taxonomic group with the best match for the alignment can be identified. The RAV can then be determined for that taxonomic group using the number of sequence reads (or votes of sequence reads) for a particular sequence group divided by the number of sequence reads identified as being bacterial, which may be for a specific region or even for a given level of a hierarchy.
A taxonomic group can include one or more bacteria and their corresponding reference sequences. A taxonomic group can correspond to any set of one or more reference sequences for one or more loci (e.g., genes) that represent the taxonomic group. Any given level of a taxonomic hierarchy would include a plurality of taxonomic groups. For instance, a reference sequence in the one group at the genus level can be in another group at the family level. A sequence read can be assigned based on the alignment to a taxonomic group when the sequence read aligns to a reference sequence of the taxonomic group. A functional group can correspond to one or more genes labeled as having a similar function. Thus, a functional group can be represented by reference sequences of the genes in the functional group, where the reference sequences of a particular gene can correspond to various bacteria. The taxonomic and functional groups can collectively be referred to as sequence groups, as each group includes one or more reference sequences that represent the group. A taxonomic group of multiple bacteria can be represented by multiple reference sequence, e.g., one reference sequence per bacteria species in the taxonomic group. Embodiments can use the degree of alignment of a sequence read to multiple reference sequences to determine which sequence group to assign the sequence read based on the alignment.
As mentioned above, a particular genomic region (e.g., gene 16S) can be analyzed. For example, the region can be amplified, and a portion of the amplified DNA fragments can be sequenced. The amplification can be to such a degree that most reads will correspond to the amplified region. Other example regions can be smaller than a gene, e.g., variable regions within a gene. The longer the region, more resolution can he obtained to determine voting to assign a sequence read to a group. Multiple non-contiguous regions can he analyzed, e.g., by amplifying multiple regions,
A. Example Determination of Relative Abundance of a Sequence Group (Feature)
As mentioned above, a relative abundance value can correspond to a proportion of sequence reads that align to at least one reference sequence of a sequence group, also referred to as a feature herein. A sequence read can he assigned to one or more sequence groups based on the alignment to the reference sequence(s) for each sequence group. A sequence read can be assigned to more than one sequence group if the assigned groups are in different categories (e.g., taxonomic or functional) or in different levels of a hierarchy (e.g., genus and family). And, a sequence group can include multiple sequences for different regions or a same region, e.g., a sequence group can include more than one base at a particular position, e.g., if the group encompasses various polymorphisms at a genomic position. A sequence group is an example of a feature that can be used to characterize a sample, e.g., when the sequence group has a statistically significant separation between the control population and the disease population.
1. Assignment to a Sequence Group
In some embodiments, sequence reads can be obtained for two ends of a nucleic acid molecule, e.g., via paired-end sequencing. Embodiments can identify whether each sequence read of a pair of sequence reads corresponds to a particular sequence group. Each sequence read can effectively have a vote, and the nucleic acid molecule can be identified as corresponding to a particular sequence group only if both sequence reads are aligned to that sequence group (alignment may allow mismatches when less than 100% sequence identity is used). In such embodiments, molecules that do not have both sequence reads aligning to the same sequence group can be discarded. The alignment to a reference sequence may be required to be perfect (i.e., no mismatches), while other embodiments can allow mismatches. Further, the alignment can be required to be unique, or else the read is discarded.
In other embodiments, a partial vote can be attributed to each sequence group to which a sequence read aligns. In one implementation, a weight of the partial vote based on the degree of alignment, e.g., whether there are any mismatches. In other implementations, each sequence read can get a vote when it does exist in a reference sequence, and that vote is weighted by the probability of its existence in humans. A total weight for a read being assigned to a particular reference sequence can be determined by various factors, each providing a weight. The total votes to the reference sequence of a group can be determined and compared to the total votes for other groups in the same level. For each read, the sequence group at a given level with the highest percentage for assignment to the read can be assigned the read. Various techniques of partial assignment can be used, e.g., Dirichlet partial assignment.
Sequencing can be advantageous for assigning sequence reads to a group, as sequencing provides the actual sequence of at least a portion of a nucleic acid molecule. The sequence might be slightly different than what has already been known for a particular taxonomic group, but it may be similar enough to assign to a particular taxonomic group. If predetermined probes were used, then that nucleic acid molecule might not be identified. Thus, one can identify unknown bacteria, but whose sequence is similar enough to an existing taxonomic group, or even assigned to an unknown group.
In some embodiments, the proportion can be the total of sequence reads, even if some are not assigned, or equivalently assigned to an unknown group. As an example, the 165 gene can be analyzed, and a read can be determined to align to one or more reference sequences in the region, e.g., with a certain number of mismatches below a threshold, but with a high enough variations to not correspond to any known taxonomic group (or functional group as discussed below). Thus, embodiments can include unassigned reads that contribute to the denominator for determining the proportion of reads of a certain sequence group relative to the sequence reads identified as being bacterial. Thus, a proportion of the bacterial population of sequence reads can be determined. Using predetermined probes would generally not allow one to identify unknown bacterial sequences.
2. Sequence Group Corresponds to a Particular Taxonomic Group
A taxonomic group can correspond to any set of one or more reference sequences for one or more loci (e.g., genes) that represent the taxonomic group. Any given level of a taxonomic hierarchy would include a plurality of taxonomic groups. The taxonomic groups of a given level of the taxonomic hierarchy would typically be mutually exclusive. Thus, a reference sequence of one taxonomic group would not be included in another taxonomic group in the same level. For example, a reference sequence in one group at the genus level would not be included in another group at the genus level. But, that reference sequence in the one group at the genus level can be in another group at the family level.
The RAV can correspond to the proportion of reads assigned to a particular taxonomic group. The proportion can be relative to various denominator values, e.g., relative to all of the sequence reads, relative to all assigned to at least one group (taxonomic or functional), or all assigned to for a given level in the hierarchy. The alignment can be implemented in any manner that can assign a sequence read to a particular taxonomic group.
For example, based on the mappings to the reference sequence(s) in the 16S region, a taxonomic group with the best match for the alignment can be identified. The RAV can then be determined for that taxonomic group using the number of sequence reads (or votes of sequence reads) for a particular sequence group divided by the number of sequence reads identified as being bacterial, which may be for a specific region or even for a given level of a hierarchy.
3. Sequence Group Corresponds to a Particular Gene or Functional Group
Instead of or in addition to determining a count of the sequence reads that correspond to a particular taxonomic group, embodiments can use a count of a number of sequence reads that correspond to a particular gene or a collection of genes having an annotation of a particular function, where the collection is called a functional group. The RAV can be determined in a similar manner as for a taxonomic group. For example, functional group can include a plurality of reference sequences corresponding to one or more genes of the functional group. Reference sequences of multiple bacteria for a same gene can correspond to a same functional group. Then, to determine the RAV, the number of sequence reads assigned to the functional group can be used to determine a proportion for the functional group.
The use of a function group, which may include a single gene, can help to identify situations where there is a small change (e.g., increase) in many taxonomic groups such that the change is too small to be statistically significant. But, the changes may all be for a same gene or set of genes of a same functional group, and thus the change for that functional group can be statistically significant, even though the changes for the taxonomic groups may not be significant. The reverse can be true of a taxonomic group being more predictive than a particular functional group, e.g., when a single taxonomic group includes many genes that have change by a relatively small amount.
As an example, if 10 taxonomic groups increase by 10%, the statistical power to discriminate between the two groups may be low when each taxonomic group is analyzed individually. But, if the increase is all for genes(s) of a same functional group, then the increase would be 100%, or a doubling of the proportion for that taxonomic group. This large increase would have a much larger statistical power for discriminating between the two groups. Thus, the functional group can act to provide a sum of small changes for various taxonomic groups. And, small changes for various functional groups, which happen to all be on a same taxonomic group, can sum to provide high statistical power for that particular taxonomic group.
The taxonomic groups and functional groups can supplement each other as the information can be orthogonal, or at least partially orthogonal as there still may be some relationship between the RAVs of each group. For example, the RAVs of one or more taxonomic groups and functional groups can be used together as multiple features of a feature vector, which is analyzed to provide a diagnosis, as is described herein. For instance, the feature vector can be compared to a disease signature as part of a characterization model.
B. Example Determination of Statistically Significant Separation of Abundance of a Sequence Group between Control and Disease Popultions
Embodiments can use the relative abundance values (RAVs) for populations of subjects that have a disease (condition population; i.e., individuals having a microbiome indicative of an oral health issue) and that do not have the disease (control population; i.e., individuals having a microbiome that is not indicative of an oral health issue). If the distribution of RAVs of a particular sequence group for the disease population is statistically different than the distribution of RAVs for the control population, then the particular sequence group can be identified for including in a disease signature. Since the two populations have different distributions, the RAV for a new sample for a sequence group in the disease signature can be used to classify (e.g., determine a probability) of whether the sample does or does not have the disease. The classification can also be used to determine a treatment, as is described herein. A discrimination level can be used to identify sequence groups that have a high predictive value. Thus, embodiment can filter out taxonomic groups that are not very accurate for providing a diagnosis.
1. Discrimination Level of Sequence Group
Once RAVs of a sequence group have been determined for the control and condition populations, various statistical tests can be used to determine the statistical power of the sequence group for discriminating between an oral health issue (condition) and no oral health issue (control). In one embodiment, the Kolmogorov-Smirnov (KS) test can be used to provide a probability value (p-value) that the two distributions are actually identical. The smaller the p-value the greater the probability to correctly identify which population a sample belongs. The larger the separation in the mean values between the two populations generally results in a smaller p-value (an example of a discrimination level). Other tests for comparing distributions can be used. The Welch's t-test presumes that the distributions are Gaussian, which is not necessarily true for a particular sequence group. The KS test, as it is a non-parametric test, is well suited for comparing distributions of taxa or functions for which the probability distributions are unknown.
The distribution of the RAVS for the control and condition populations can be analyzed to identify sequence groups with a large separation between the two distributions. The separation can be measured as a p-value (See example section). For example, the relative abundance values for the control population may have a distribution peaked at a first value with a certain width and decay for the distribution. And, the disease population can have another distribution that is peaked a second value that is statistically different than the first value. In such an instance, an abundance value of a control sample has a lower probability to be within the distribution of abundance values encountered for the disease samples. The larger the separation between the two distributions, the more accurate the discrimination is for determining whether a given sample belongs to the control population or the disease population. As is discussed later, the distributions can be used to determine a probability for an RAV as being in the control population and determine a probability for the RAV being in the disease population.
Similarly,
Similarly,
2. Prevalence of Sequence Group in Population
In some embodiments, certain samples may not have any presence of a particular taxonomic group, or at least not a presence above a relatively low threshold (i.e., a threshold below either of the two distributions for the control and condition population). Thus, a particular sequence group may be prevalent in the population, e.g., more than 30% of the population may have the taxonomic group. Another sequence group may be less prevalent in the population, e.g., showing up in only 5% of the population. The prevalence (e.g., percentage of population) of a certain sequence group can provide information as to how likely the sequence group may be used to determine a diagnosis.
In such an example, the sequence group can be used to determine a status of the disease (e.g., diagnose for the disease) when the subject falls within the 30%. But, when the subject does not fall within the 30%, such that the taxonomic group is simply not present, the particular taxonomic group may not be helpful in determining a diagnosis of the subject. Thus, whether a particular taxonomic group or functional group is useful in diagnosing a particular subject can be dependent on whether nucleic acid molecules corresponding to the sequence group are actually sequenced.
Accordingly, the disease signature can include more sequence groups that are used for a given subject. As an example, the disease signature can include 100 sequence groups, but only 60 of sequence groups may be detected in a sample. The classification of the subject (including any probability for being in the application) would be determined based on the 60 sequence groups.
C. Example Generation of Characterization Model
The sequence groups with high discrimination levels (e.g., low p-values) for a given condition (e.g., an oral health issue) can be identified and used as part of a characterization model, e.g., which uses a disease signature to determine a probability of a subject having the disease. The disease signature can include a set of sequence groups as well as discriminating criteria (e.g., cutoff values and/or probability distributions) used to provide a classification of the subject. The classification can be binary (e.g., indicative of an oral health issue or not indicative of an oral health issue) or have more classifications (e.g., probability of being indicative of an oral health issue or not being indicative of an oral health issue). Which sequence groups of the disease signature that are used in making a classification be dependent on the specific sequence reads obtained, e.g., a sequence group would not be used if no sequence reads were assigned to that sequence group. In some embodiments, a separate characterization model can be determined for different populations, e.g., by geography where the subject is currently residing (e.g., country, region, or continent), the generic history of the subject (e.g., ethnicity), or other factors.
1. Selection of Sequence Groups
As mentioned above, sequence groups having at least a specified discrimination level can be selected for inclusion in the characterization model. In various embodiments, the specified discrimination level can be an absolute level (e.g., having a p-value below a specified value), a percentage (e.g., being in the top 10% of discriminating levels), or a specified number of the top discrimination levels (e.g., the top 100 discriminating levels). In some embodiments, the characterization model can include a network graph, where each node in a graph corresponds to a sequence group having at least a specified discrimination level.
The sequence groups used in a disease signature of a characterization model can also be selected based on other factors. For example, a particular sequence group may only be detected in a certain percentage of the population, referred to as a coverage percentage. An ideal sequence group would be detected in a high percentage of the population and have a high discriminating level (e.g., a low p-value). A minimum percentage may be required before adding the sequence group to the characterization model for a particular disease e.g., an oral health issue). The minimum percentage can vary based on the accompanying discriminating level. For instance, a lower coverage percentage may be tolerated if the discriminating level is higher. As a further example, 95% of the patients with a disease may be classified with one or a combination of a few sequence groups, and the 5% remaining can be explained based on one sequence group, which relates to the orthogonality or overlap between the coverage of sequence groups. Thus, a sequence group that provides discriminating power for 5% of the individuals having the disease (e.g., an oral health issue) may be valuable.
Another factor for determining which sequence to include in a disease signature of the characterization model is the overlap in the subjects exhibiting the sequence groups of a disease signature. For example, to sequence groups can both have a high coverage percentage, but sequence groups may cover the exact same subjects. Thus, adding one of the sequence groups does increase the overall coverage of the disease signature. In such a situation, the two sequence groups can be considered parallel to each other. Another sequence group can be selected to add to the characterization model based on the sequence group covering different subjects than other sequence groups already in the characterization model. Such a sequence group can be considered orthogonal to the already existing sequence groups in the characterization model.
As examples, selecting a sequence group may consider the following factors. A taxa may appear in 100% of control individuals and in 100% of individuals having a specified disease (e.g., an oral health issue), but where the distributions are so close in both groups, that knowing the relative abundance of that taxa only allows to catalogue a few individuals as having the disease or lacking the disease (i.e it has a low discriminating level). Whereas, a taxa that appears in only 20% of individuals not having the disease and 30% of individuals having the disease can have distributions of relative abundance that are so different from one another, it allows to catalogue 20% of individuals not having the disease and 30% of individuals having the disease (i.e. it has a high discriminating level).
In some embodiments, machine learning techniques can allow the automatic identification of the best combination of features (e.g., sequence groups). For instance, a Principal Component Analysis can reduce the number of features used for classification to only those that are the most orthogonal to each other and can explain most of the variance in the data. The same is true for a network theory approach, where one can create multiple distance metrics based on different features and evaluate which distance metric is the one that best separates individuals having the disease (an oral health issue) from individuals that do not have the disease.
2. Discrimination Criteria Sequence Groups
The discrimination criteria for the sequence groups included in the disease signature of a characterization model can be determined based on the disease distributions and the control distributions for the disease. For example, a discrimination criterion for a sequence group can be a cutoff value that is between the mean values for the two distributions. As another example, discrimination criteria for a sequence group can include probability distributions for the control and disease populations. The probability distributions can be determined in a separate manner from the process of determining the discrimination level.
The probability distributions can be determined based on the distribution of RAVs for the two populations. The mean values (or other average or median) for the two populations can be used to center the peaks of the two probability distributions. For example, if the mean RAV of the disease population is 20% (or 0.2), then the probability distribution for the disease population can have its peak at 20%. The width or other shape parameters (e.g., the decay) can also be determined based on the distribution of RAVs for the disease population. The same can be done for the control population.
D. Use of Sequence Groups
The sequence groups included in the disease signature of the characterization can be used to classify a new subject. The sequence groups can be considered features of the feature vector, or the RAVs of the sequence groups considered as features of a feature vector, where the feature vector can be compared to the discriminating criteria of the disease signature. For instance, the RAVs of the sequence groups for the new subject can be compared to the probability distributions for each sequence group of the disease signature. If an RAV is zero or nearly zero, then the sequence group may be skipped and not used in the classification.
The RAVs for sequence groups that are exhibited in the new subject can be used to determine the classification. For example, the result (e.g., a probability value) for each exhibited sequence group can be combined to arrive at the final classification. As another example, clustering of the RAVs can be performed, and the clusters can be used to determine a classification of a disease.
1. Classification of Disease Using Sequence Groups
Embodiments can provide a method for determining a classification of the presence or absence for a disease and/or determine a course of treatment for an individual human having the disease (an oral health issue such as dental decay or gingivitis). The method can be performed by a computer system, as described herein.
In block 20, sequence reads of bacterial DNA obtained from analyzing a test sample from the individual human are received. The analysis can be done with various techniques, e.g., as described herein, such as sequencing or hybridization arrays. The sequence reads can he received at a computer system, e.g., from a detection apparatus, such as a sequencing machine that provides data to a storage device (which can be loaded into the computer system) or across a network to the computer system.
In block 21, the sequence reads are mapped to a bacterial sequence database to obtain a plurality of mapped sequence reads. The bacterial sequence database includes a plurality of reference sequences of a plurality of bacteria. The reference sequences can be for predetermined region(s) of the bacteria, e.g., the 16S region.
In block 22, the mapped sequence reads are assigned to sequence groups based on the mapping to obtain assigned sequence reads assigned to at least one sequence group. A sequence group includes one or more of the plurality of reference sequences. The mapping can involve the sequence reads being mapped to one or more predetermined regions of the reference sequences. For example, the sequence reads can be mapped to the 16S gene. Thus, the sequence reads do not have to be mapped to the whole genome, but only to the region(s) covered by the reference sequences of a sequence group.
In block 23, a total number of assigned sequence reads is determined. In some embodiments, the total number of assigned reads can include reads identified as being bacterial, but not assigned to a known sequence group. In other embodiments, the total number can be a sum of sequence reads assigned to known sequence groups, where the sum may include any sequence read assigned to at least one sequence group.
In block 24, relative abundance value(s) can be determined. For example, for each sequence group of a disease signature set of one or more sequence groups selected from TABLE A, a relative abundance value of assigned sequence reads assigned to the sequence group relative to the total number of assigned sequence reads can he determined. The relative abundance values can form a test feature vector, where each values of the test feature vector is an RAV of a different sequence group.
In block 25, the test feature vector is compared to calibration feature vectors generated from relative abundance values of calibration samples having a known status of the disease. The calibration samples may be samples of a disease population and samples of a control population. In some embodiments, the comparison can involve various machine learning techniques, such as supervised machine learning (e.g. decision trees, nearest neighbor, support vector machines, neural networks, naïve Bayes classifier, etc . . . ) and unsupervised machine learning (e.g., clustering, principal component analysis, etc . . . ).
In one embodiment, clustering can use a network approach, where the distance between each pair of samples in the network is computed based on the relative abundance of the sequence groups that are relevant for each disease. Then, a new sample can be compared to all samples in the network, using the same metric based on relative abundance, and it can be decided to which cluster it should belong. A meaningful distance metric would allow all individuals having the disease (an oral health issue) to form one or a few clusters and all individuals lacking the disease to form one or a few clusters. One distance metric is the Bray-Curtis dissimilarity, or equivalently a similarity network, where the metric is 1—Bray-Curtis dissimilarity. Another example distance metric is the Tanimoto coefficient.
In some embodiments, the feature vectors may be compared by transforming the RAVs into probability values, thereby forming probability vectors. Similar processing for the feature vectors can be performed for the probability, with such a process still involving a comparison of the feature vectors since the probability vectors are generated from the feature vectors.
Block 26 can determine a classification of the presence or absence of the disease (e.g., an oral health issue) and/or determine a course of treatment for an individual human having the disease based on the comparing. For example, the cluster to which the test feature vector is assigned may be a disease cluster, and the classification can be made that the individual human has the disease or a certain probability for having the disease.
In one embodiment involving clustering, the calibration feature vectors can be clustered into a control cluster not having the disease and a disease cluster having the disease. Then, which cluster the test feature vector belongs can he determined. The identified cluster can he used to determine the classification or select a course of treatment. In one implementation, the clustering can use a Bray-Curtis dissimilarity.
In one embodiment involving a decision tree, the comparison may be performed to by comparing the test feature vector to one or more cutoff values e.g., as a corresponding cutoff vector), where the one or more cutoff values are determined from the calibration feature vectors, thereby providing the comparison. Thus, the comparing can include comparing each of the relative abundance values of the test feature vector to a respective cutoff value determined from the calibration feature vectors generated from the calibration samples. The respective cutoff values can be determined to provide an optimal discrimination for each sequence group.
2. Use of Probability Values
A new sample can be measured to detect the RAVs for the sequence groups in the disease signature. The RAV for each sequence group can be compared to the probability distributions for the control and disease populations for the particular sequence group. For example, the probability distribution for the disease population can provide an output of a probability (disease probability) of having the disease for a given input of the RAV. Similarly, the probability distribution for the control population can provide an output of a probability (control probability) of not having the disease for a given input of the RAV. Thus, the value of the probability distribution at the RAV can provide the probability of the sample being in each of the populations. Thus, it can be determined which population the sample is more likely to belong to, by taking the maximum probability.
In some embodiments, just the maximum probability is used in further steps of a characterization process. In other embodiments, both the disease probability and the control probability are used. As noted above, the probability distributions used here for classification may be different than the statistical test used to determine whether the distribution of RAV values are separated, e.g., the KS test.
A total probability across sequence groups of a disease signature can be used. For all of the sequence groups that are measured, a disease probability can be determined for whether the sample is in the disease group and a control probability can be determined for whether the sample is in the control population. In other embodiments, just the disease probabilities or just the control probabilities can be determined.
The probabilities across the sequence groups can be used to determine a total probability. For example, an average of the disease probabilities can be determined, thereby obtaining a final disease probability of the subject having the disease based on the disease signature. An average of the control probabilities can be determined, thereby obtaining a final control probability of the subject not having the disease based on the disease signature.
In one embodiment, the final disease probability and final control probability can be compared to each other to determine the final classification. For instance, a difference between the two final probabilities can be determined, and a final classification probability determined from the difference. A large positive difference with final disease probability being higher would result in a higher final classification probability of the subject having the disease.
In other embodiments, only the final disease probability can be used to determine the final classification probability. For example, the final classification probability can be the final disease probability. Alternatively, the final classification probability can be one minus the final control probability, or 100% minus the final control probability depending on the formatting of the probabilities.
In some embodiments, a final classification probability for one disease of a class can be combined with other final classification probabilities of other disease of the same class. The aggregated probability can then be used to determine whether the subject has at least one of the class of diseases. Thus, embodiments can determine whether a subject has a health issue that may include a plurality of diseases associated with that health issue.
The classification can be one of the final probabilities. In other examples, embodiments can compare a final probability to a threshold value to make a determination of whether the disease exists. For example, the respective disease probabilities can be averaged, and an average can be compared to a threshold value to determine whether the disease exists. As another example, the comparison of the average to the threshold value can provide a treatment for treating the subject.
Described herein, and with reference to the FIGS, are additional illustrative embodiments of the methods, compositions, and systems provided herein. It will be appreciated that one of ordinary skill in the art can readily determine where and when any one or more of the methods, compositions, and/or systems described above can be utilized additionally, or alternatively, in the embodiments described below.
As shown in
The first method 100 functions to generate models that can be used to characterize and/or diagnose subjects according to at least one of their microbiome composition and functional features (e.g., as a clinical diagnostic, as a companion diagnostic, etc.), and provide therapeutic measures (e.g., probiotic-based therapeutic measures, phage-based therapeutic mea.sures, small-molecule-based therapeutic measures, prebiotic-based therapeutic measures, clinical measures, etc.) to subjects based upon rnicrobiorne analysis for a population of subjects. As such, data from the population of subjects can be used to characterize subjects according to their microbiome composition and/or functional features, indicate states of health and areas of improvement based upon the characterization(s), and promote one or more therapies that can modulate the composition of a subject's microbiome toward one or more of a set of desired equilibrium states.
In variations, the method 100 can be used to promote targeted therapies to subjects having a microbiome indicative of an oral health issue. In some cases, the targeted therapies are promoted when the oral health issue produces observed differences in dental decay or gingivitis or at least one of social behavior, motor behavior, and energy levels, gastrointestinal heath, etc. In these variations, diagnostics associated with an oral health issue can be typically assessed using one or more of: a survey instrument, a dental exam, and any other standard tool. As such, the method 100 can be used to characterize the effects of an oral health issue, including disorders, and/or adverse states in an entirely non-typical method. In particular, the inventors propose that characterization of the microbiome of individuals can be useful for predicting the likelihood of an oral health issue in subjects. Such characterizations can also be useful for screening for symptoms related to an oral health issue and/or determining a course of treatment for an individual human having a microbiome indicative of an oral health issue. For example, by deep sequencing bacterial DNAs from subjects having an oral health issue and control subjects, the inventors propose that features associated with certain microbiome compositional and/or functional features (e.g., the amount of certain bacteria and/or bacterial sequences corresponding to certain genetic pathways) can be used to predict the presence or absence of a microbiome indicative of an oral health issue. The bacteria and genetic pathways in some cases are present in a certain abundance in individuals having a microbiome indicative of an oral health issue as discussed in more detail below whereas the bacteria and genetic pathways are at a statistically different abundance in individuals not having a microbiome indicative of an oral health issue.
As such, in some embodiments, outputs of the first method 100 can be used to generate diagnostics and/or provide therapeutic measures for a subject based upon an analysis of the subject's microbiome composition and/or functional features of the subject's microbiome. Thus, as shown in
Thus, methods 100 and/or 200 can function to generate models that can be used to classify individuals and/or provide therapeutic measures (e.g., therapy recommendations, therapies, therapy regimens, etc.) to individuals based upon microbiome analysis for a population of individuals. As such, data from the population of individuals can be used to generate models that can classify individuals according to their microbiome compositions (e.g., as a diagnostic measure), indicate states of health and areas of improvement based upon the classification(s), and/or provide therapeutic measures that can push the composition of an individual's microbiome toward one or more of a set of improved equilibrium states. Variations of the second method 200 can further facilitate monitoring and/or adjusting of therapies provided to an individual, for instance, through reception, processing, and analysis of additional samples from an individual throughout the course of therapy.
In one application, at least one of the methods 100, 200 is implemented, at least in part, at a system 300, as shown in
Thus, the methods 100, 200 can be implemented for a population of subjects (e.g., including the subject, excluding the subject), wherein the population of subjects can include subjects dissimilar to and/or similar to the subject health condition, in dietary needs, in demographic features, etc.). Thus, information derived from the population of subjects can be used to provide additional insight into connections between behaviors of a subject and effects on the subject's microbiome, due to aggregation of data from a population of subjects.
A. Sample Handling
Block S110 recites: receiving an aggregate set of biological samples from a population of subjects, which functions to enable generation of data from which models for characterizing subjects and/or providing therapeutic measures to subjects can be generated. In Block S110, biological samples are preferably received from subjects of the population of subjects in a non-invasive manner. In variations, non-invasive manners of sample reception can use any one or more of: a permeable substrate (e.g., a swab configured to wipe a region of a subject's body, toilet paper, a sponge, etc.), a non-permeable substrate (e.g., a slide, tape, etc.), a container (e.g., vial, tube, bag, etc.) configured to receive a sample from a region of a subject's body, and any other suitable sample-reception element. In a specific example, samples can be collected from one or more of a subject's nose, skin, genitals, mouth, and gut in a non-invasive manner (e.g., using a swab and a vial). However, one or more biological samples of the set of biological samples can additionally or alternatively be received in a semi-invasive manner or an invasive manner. In variations, invasive manners of sample reception can use any one or more of: a needle, a syringe, a biopsy element, a lance, and any other suitable instrument for collection of a sample in a semi-invasive or invasive manner. In specific examples, samples can comprise blood samples, pla.sma/serum samples (e.g., to enable extraction of cell-free DNA), cerebrospinal fluid, and tissue samples. In some cases, the sample is a stool sample, or a sampl ucleic acid sample, such as a DNA sample) extracted from a stool sample.
In the above variations and examples, samples can be taken from the bodies of subjects without facilitation by another entity (e.g., a caretaker associated with an individual, a health care professional, an automated or semi-automated sample collection apparatus, etc.), or can alternatively be taken from bodies of individuals with the assistance of another entity. In one example, wherein samples are taken from the bodies of subjects without facilitation by another entity in the sample extraction process, a sample-provision kit can be provided to a subject. In the example, the kit can include one or more swabs or sample vials for sample acquisition, one or more containers configured to receive the swab(s) or sample vials for storage, instructions for sample provision and setup of a user account, elements configured to associate the sample(s) with the subject (e.g., barcode identifiers, tags, etc.), and a receptacle that allows the sample(s) from the individual to be delivered to a sample processing operation (e.g., by a mail delivery system). In another example, wherein samples are extracted from the user with the help of another entity, one or more samples can be collected in a clinical or research setting from a subject (e.g., during a clinical appointment).
In Block S110, the aggregate set of biological samples is preferably received from a wide variety of subjects, and can involve samples from human subjects and/or non-human subjects. In relation to human subjects, Block S110 can include receiving samples from a wide variety of human subjects, collectively including subjects of one or more of: different demographics (e.g., genders, ages, marital statuses, ethnicities, nationalities, socioeconomic statuses, sexual orientations, etc.), different health conditions (e.g., health and disease states), different living situations (e.g., living alone, living with pets, living with a significant other, living with children, etc.), different dietary habits (e.g., omnivorous, vegetarian, vegan, sugar consumption, acid consumption, etc.), different behavioral tendencies (e.g., levels of physical activity, drug use, alcohol use, etc.), different levels of mobility (e.g., related to distance traveled within a given time period), biomarker states (e.g., cholesterol levels, lipid levels, etc.), weight, height, body mass index, genotypic factors, and any other suitable trait that has an effect on microbiome composition. As such, as the number of subjects increases, the predictive power of feature-based models generated in subsequent blocks of the method 100 increases, in relation to characterizing a variety of subjects based upon their microbiomes. Additionally or alternatively, the aggregate set of biological samples received in Block S110 can include receiving biological samples from a targeted group of similar subjects in one or more of: demographic traits, health conditions, living situations, dietary habits, behavior tendencies, levels of mobility, age range (e.g., pediatric, adulthood, geriatric), and any other suitable trait that has an effect on microbiome composition. Additionally or alternatively, the methods 100, and/or 200 can be adapted to characterize diseases typically detected by way of lab tests polymerase chain reaction based tests, cell culture based tests, blood tests, biopsies, chemical tests, etc.), physical detection methods (e.g., manometry), medical history based assessments, behavioral assessments, and itnagenology based assessments. Additionally or alternatively, the methods 100, 200 can be adapted to characterization of acute conditions, chronic conditions, conditions with difference in prevalence for different demographics, conditions having characteristic disease areas (e.g., the head, the gut, endocrine system diseases, the heart, nervous system diseases, respiratory diseases, immune system diseases, circulatory system diseases, renal system diseases, locomotor system diseases, etc.), and comorbid conditions.
In some embodiments, receiving the aggregate set of biological samples in Block S110 can be performed according to embodiments, variations, and examples of sample reception as described in U.S. application Ser. No. 14/593,424 filed on 9, Jan. 2015 and entitled “Method and System for Microbiome Analysis”, which is incorporated herein in its entirety by this reference. However, receiving the aggregate set of biological samples in Block S110 can additionally or alternatively be performed in any other suitable manner. Furthermore, some alternative variations of the first method 100 can omit Block S110, with processing of data derived from a set of biological samples performed as described below in subsequent blocks of the method 100.
B. Sample Analysis
Block S120 recites: characterizing a microbiome composition and/or functional features for each of the aggregate set of biological samples associated with a population of subjects, thereby generating at least one of a microbiome composition dataset and a microbiome functional diversity dataset for the population of subjects. Block S120 functions to process each of the aggregate set of biological samples, in order to determine compositional and/or functional aspects associated with the microbiome of each of a population of subjects. Compositional and functional aspects can include compositional aspects at the microorganism level, including parameters related to distribution of microorganisms across different groups of kingdoms, phyla, classes, orders, families, genera, species, subspecies, strains, infraspecies taxon (e.g., as measured in total abundance of each group, relative abundance of each group, total number of groups represented, etc.), and/or any other suitable taxa. Compositional and functional aspects can also be represented in terms of operational taxonomic units (OTUs). Compositional and functional aspects can additionally or alternatively include compositional aspects at the genetic level (e.g., regions determined by multilocus sequence typing, 16S sequences, 18S sequences, ITS sequences, other genetic markers, other phylogenetic markers, etc.). Compositional and functional aspects can include the presence or absence or the quantity of genes associated with specific functions (e.g., enzyme activities, transport functions, immune activities, etc.). Outputs of Block S120 can thus be used to provide features of interest for the characterization process of Block S140, wherein the features can be microorganism-based (e.g., presence of a genus of bacteria), genetic-based (e.g., based upon representation of specific genetic regions and/or sequences) and/or functional-based (e.g., presence of a specific catalytic activity, presence of metabolic pathways, etc.).
In one variation, Block S120 can include characterization of features based upon identification of phylogenetic markers derived from bacteria and/or archaea in relation to gene families associated with one or more of: ribosomal protein S2, ribosomal protein S3, ribosomal protein S5, ribosomal protein S7, ribosomal protein S8, ribosomal protein S9, ribosomal protein S10, ribosomal protein S11, ribosomal protein S12/S23, ribosomal protein S13, ribosomal protein S15P/S13e, ribosomal protein S17, ribosomal protein S19, ribosomal protein L1, ribosomal protein L2, ribosomal protein L3, ribosomal protein L4/L1e, ribosomal protein L5, ribosomal protein L6, ribosomal protein L10, ribosomal protein L11, ribosomal protein L13, ribosomal protein L14b/L23e, ribosomal protein L15, ribosomal protein L16/L10E, ribosomal protein L18P/L5E, ribosomal protein L22, ribosomal protein L24, ribosomal protein L25/L23, ribosomal protein L29, translation elongation factor EF-2, translation initiation factor IF-2, metalloendopeptidase, ffh signal recognition particle protein, phenylalanyl-tRNA synthetase alpha subunit, phenylalanyl- tRNA synthetase beta subunit, tRNA pseudouridine synthase B, porphobilinogen deaminase, phosphoribosylformylglycinamidine cyclo-ligase, and ribonuclease HI1. However, the markers can include any other suitable marker(s).
Characterizing the microbiome composition and/or functional features for each of the aggregate set of biological samples in Block S120 thus can include a combination of sample processing techniques (e.g., wet laboratory techniques) and computational techniques (e.g., utilizing tools of bioinformatics) to quantitatively and/or qualitatively characterize the microbiorne and functional features associated with each biological sample from a subject or population of subjects.
In variations, sample processing in Block S120 can include any one or more of: lysing a biological sample, disrupting membranes in cells of a biological sample, separation of undesired elements (e.g., RNA, proteins) from the biological sample, purification of nucleic acids (e.g., DNA) in a biological sample, amplification of nucleic acids from the biological sample, further purification of amplified nucleic acids of the biological sample, and sequencing of amplified nucleic acids of the biological sample. Thus, portions of Block S120 can be implemented using embodiments, variations, and examples of the sample handling network and/or computing system as described in U.S. application Ser. No. 14/593,424 filed on 9, Jan. 2015 and entitled “Method and System for microbiome Analysis”, which is incorporated herein in its entirety by this reference. Thus the computing system implementing one or more portions of the method 100 can be implemented in one or more computing systems, wherein the computing system(s) can be implemented at least in part in the cloud and/or as a machine (e.g., computing machine, server, mobile computing device, etc.) configured to receive a computer-readable medium storing computer-readable instructions. However, Block S120 can be performed using any other suitable system(s).
In variations, lysing a biological sample and/or disrupting membranes in cells of a biological sample preferably includes physical methods (e.g., bead beating, nitrogen decompression, homogenization, sonication), which omit certain reagents that produce bias in representation of certain bacterial groups upon sequencing. Additionally or alternatively, lysing or disrupting in Block S120 can involve chemical methods (e.g., using a detergent, using a solvent, using a surfactant, etc.). Additionally or alternatively, lysing or disrupting in Block S120 can involve biological methods. In variations, separation of undesired elements can include removal of RNA using RNases and/or removal of proteins using proteases. In variations, purification of nucleic acids can include one or more of: precipitation of nucleic acids from the biological samples (e.g., using alcohol-based precipitation methods), liquid- liquid based purification techniques (e.g., phenol-chloroform extraction), chromatography-based purification techniques (e.g., column adsorption), purification techniques involving use of binding moiety-bound particles (e.g., magnetic beads, buoyant beads, beads with size distributions, ultrasonically responsive beads, etc.) configured to bind nucleic acids and configured to release nucleic acids in the presence of an elution environment e.g., having an elution solution, providing a pH shift, providing a temperature shift, etc.), and any other suitable purification techniques.
In variations, performing an amplification operation S123 on purified nucleic acids can include performing one or more of: polymerase chain reaction (PCR)-based techniques (e.g., solid-phase PCR, RT-PCR, qPCR, multiplex PCR, touchdown PCR, nanoPCR, nested PCR, hot start PCR, etc.), helicase-dependent amplification (HDA), loop mediated isothermal amplification (LAMP), self-sustained sequence replication (3SR), nucleic acid sequence based amplification (NASBA), strand displacement amplification (SDA), rolling circle amplification (RCA), ligase chain reaction (LCR), and any other suitable amplification technique. In amplification of purified nucleic acids, the primers used are preferably selected to prevent or minimize amplification bias, as well as configured to amplify nucleic acid regions/sequences (e.g., of the 16S region, the 18S region, the ITS region, etc.) that are informative taxonomically, phylogenetically, for diagnostics, for formulations (e.g., for probiotic formulations), and/or for any other suitable purpose. Thus, universal primers (e.g., a F27-R338 primer set for 165 rRNA, a F515-R806 primer set for 165 rRNA, etc.) configured to avoid amplification bias can be used in amplification. Primers used in variations of Block S120 (e.g., S123 and/or S124) can additionally or alternatively include incorporated barcode sequences specific to each biological sample, which can facilitate identification of biological samples post-amplification. Primers used in variations of Block S120 (e.g., S123 and/or S124) can additionally or alternatively include adaptor regions configured to cooperate with sequencing techniques involving complementary adaptors (e.g., according to protocols for Illumina Sequencing).
Identification of a primer set for a multiplexed amplification operation can be performed according to embodiments, variations, and examples of methods described in U.S. App. No. 62/206,654 filed 18, Aug. 2015 and entitled “Method and. System for Multiplex Primer Design”, which is herein incorporated in its entirety by this reference. Performing a multiplexed amplification operation using a set of primers in Block S123 can additionally or alternatively be performed in any other suitable manner.
Additionally or alternatively, as shown in
In a specific example, amplification and sequencing of nucleic acids from biological samples of the set of biological samples includes: solid-phase PCR involving bridge amplification of DNA fragments of the biological samples on a substrate with oligo adapters, wherein amplification involves primers having a forward index sequence (e.g., corresponding to an illumina forward index for miSeq/NextSeq/HiSeq platforms) and/or a reverse index sequence (e.g., corresponding to an Illumina reverse index for MiSeq/NextSeq/HiSeq platforms), a forward barcode sequence and/or a reverse barcode sequence, optionally a transposase sequence (e.g., corresponding to a transposase binding site for MiSeq/NextSeq/HiSeq platforms), optionally a linker (e.g., a zero, one, or two-base fragment configured to reduce homogeneity and improve sequence results), optionally an additional random base, and optionally a sequence for targeting a specific target region (e.g., 16S region, 18S region, ITS region). In some cases, amplification involves one or both primers having any combination of the foregoing elements, or all of the foregoing elements. Amplification and sequencing can further be performed on any suitable amplicaon, as indicated throughout the disclosure. In the specific example, sequencing comprises Illumina sequencing (e.g., with a HiSeq platform, with a MiSeq platform, with a NextSeq platform, etc.) using a sequencing-by-synthesis technique. Additionally or alternatively, any other suitable next generation sequencing technology (e.g., PacBio platform, MinION platform, Oxford Nanopore platform, etc.) can be used. Additionally or alternatively, any other suitable sequencing platform or method can be used (e.g., a Roche 454 Life Sciences platform, a Life Technologies SOLiD platform, etc.). In examples, sequencing can include deep sequencing to quantify the number of copies of a particular sequence in a sample and then also be used to determine the relative abundance of different sequences in a sample. The sequencing depth can be, or be at least about 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55 ,56, 57, 58, 59, 60, 70, 80, 90, 100, 110, 120, 130, 150, 200, 300, 500, 500, 700, 1000, 2000, 3000, 4000, 5000 or more.
Some variations of sample processing in Block S120 can include further purification of amplified nucleic acids (e.g., PCR products) prior to sequencing, which functions to remove excess amplification elements (e.g., primers, dUTPs, enzymes, salts, etc.). In examples, additional purification can be facilitated using any one or more of: purification kits, buffers, alcohols, pH indicators, chaotropic salts, nucleic acid binding filters, centrifugation, and any other suitable purification technique.
In variations, computational processing in Block S120 can include any one or more of: performing a sequencing analysis operation S124 including identification of microbiome-derived sequences (e.g., as opposed to subject sequences and contaminants), performing an alignment and/or mapping operation S125 of microbiome-derived sequences (e.g., alignment of fragmented sequences using one or more of single-ended alignment, ungapped alignment alignment, gapped alignment, pairing), and generating features S126 derived from compositional and/or functional aspects of the microbiome associated with a biological sample.
Performing the sequencing analysis operation S124 with identification of microbiome-derived sequences can include mapping of sequence data from sample processing to a subject reference genome (e.g., provided by the Genome Reference Consortium), in order to remove subject genome-derived sequences. Unidentified sequences remaining after mapping of sequence data to the subject reference genome can then be further clustered into operational taxonomic units (OTUs) based upon sequence similarity and/or reference-based approaches (e.g., using VAMPS, using MG-RAST, and/or using QIIME databases), aligned (e.g., using a genome hashing approach, using a Needleman-Wunsch algorithm, using a Smith-Waterman algorithm), and mapped to reference bacterial genomes (e.g., provided by the National Center for Biotechnology Information), using an alignment algorithm (e.g., Basic Local Alignment Search Tool, FPGA accelerated alignment tool, BWT-indexing with BWA, BWT-indexing with SOAP, BWT-indexing with Bowtie, etc.). Mapping of unidentified sequences can additionally or alternatively include mapping to reference archaeal genomes, viral genomes and/or eukaryotic genomes. Furthermore, mapping of taxa can be performed in relation to existing databases, and/or in relation to custom-generated databases.
Additionally or alternatively, in relation to generating a microbiome functional diversity dataset, Block S120 can include extracting candidate features associated with functional aspects of one or more microbiome components of the aggregate set of biological samples S127, as indicated in the microbiome composition dataset. Extracting candidate functional features can include identifying functional features associated with one or more of: prokaryotic clusters of orthologous groups of proteins (COGs); eukaryotic clusters of orthologous groups of proteins (KOGs); any other suitable type of gene product; an RNA processing and modification functional classification; a chromatin structure and dynamics functional classification; an energy production and conversion functional classification; a cell cycle control and mitosis functional classification; an amino acid metabolism and transport functional classification; a nucleotide metabolism and transport functional classification; a carbohydrate metabolism and transport functional classification; a coenzyme metabolism functional classification; a lipid metabolism functional classification; a translation functional classification; a transcription functional classification; a replication and repair functional classification; a cell wall/membrane/envelop biogenesis functional classification; a cell motility functional classification; a post-translational modification, protein turnover, and chaperone functions functional classification; an inorganic ion transport and metabolism functional classification; a secondary metabolites biosynthesis, transport and catabolism functional classification; a signal transduction functional classification; an intracellular trafficking and secretion functional classification; a nuclear structure functional classification; a cytoskeleton functional classification; a general functional prediction only functional classification; and a function unknown functional classification; and any other suitable functional classification.
Additionally or alternatively, extracting candidate functional features in Block S127 can include identifying functional features associated with one or more of: systems information (e.g., pathway maps for cellular and organismal functions, modules or functional units of genes, hierarchical classifications of biological entities); genomic information (e.g., complete genomes, genes and proteins in the complete genomes, orthologous groups of genes in the complete genomes); chemical information (e.g., chemical compounds and glycans, chemical reactions, enzyme nomenclature); health information (e.g., human diseases, approved drugs, crude drugs and health-related substances); metabolism pathway maps; genetic information processing (e.g., transcription, translation, replication and repair, etc.) pathway maps; environmental information processing (e.g., membrane transport, signal transduction, etc.) pathway maps; cellular processes (e.g., cell growth, cell death, cell membrane functions, etc.) pathway maps; organismal systems (e.g., immune system, endocrine system, nervous system, etc.) pathway maps; human disease pathway maps; drug development pathway maps; and any other suitable pathway map.
In extracting candidate functional features, Block S127 can comprise performing a search of one or more databases, such as the Kyoto Encyclopedia of Genes and Genomes (KEGG) and/or the Clusters of Orthologous Groups (COGs) database managed by the National Center for Biotechnology Information (NCBI). Searching can be performed based upon results of generation of the microbiome composition dataset from one or more of the set of aggregate biological samples and/or sequencing of material from the set of samples. In more detail, Block S127 can include implementation of a data-oriented entry point to a KEGG database including one or more of a KEGG pathway tool, a KEGG BRITE tool, a KEGG module tool, a KEGG ORTHOLOGY (KO) tool, a KEGG genome tool, a KEGG genes tool, a KEGG compound tool, a KEGG glycan tool, a KEGG reaction tool, a KEGG disease tool, a KEGG drug tool, or a KEGG medicus tool. Searching can additionally or alternatively be performed according to any other suitable filters. Additionally or alternatively, Block S127 can include implementation of an organism-specific entry point to a KEGG database including a KEGG organisms tool. Additionally or alternatively, Block S127 can include implementation of an analysis tool including one or more of: a KEGG mapper tool that maps KEGG pathway, BRITE, or module data; a KEGG atlas tool for exploring KEGG global maps, a BlastKOALA tool for genome annotation and KEGG mapping, a BLAST/FASTA sequence similarity search tool, a SIMCOMP chemical structure similarity search tool, and a SUBCOMP chemical substructure search tool. In specific examples, Block S127 can include extracting candidate functional features, based on the microbiome composition dataset, from a KEGG database resource and a COG database resource; moreover, Block S127 can comprise extracting candidate functional features in any other suitable manner. For instance, Block S127 can include extracting candidate functional features, including functional features derived from a Gene Ontology functional classification, and/or any other suitable features.
In one example, a taxonomic group can include one or more bacteria and their corresponding reference sequences. A sequence read can be assigned based on the alignment to a taxonomic group when the sequence read aligns to a reference sequence of the taxonomic group. A functional group can correspond to one or more genes labeled as having a similar function. Thus, a functional group can be represented by reference sequences of the genes in the functional group, where the reference sequences of a particular gene can correspond to various bacteria. The taxonomic and functional groups can collectively be referred to as sequence groups, as each group includes one or more reference sequences that represent the group. A taxonomic group of multiple bacteria can be represented by multiple reference sequence, e.g., one reference sequence per bacteria species in the taxonomic group. Embodiments can use the degree of alignment of a sequence read to multiple reference sequences to determine which sequence group to assign the sequence read based on the alignment.
1. Analysis of Sequence Groups
Instead of or in addition to determining a count of the sequence reads that correspond to a particular taxonomic group, embodiments can use a count of a number of sequence reads that correspond to a particular gene or a collection of genes having an annotation of a. particular function, where the collection is called a functional group. The RAV can be determined in a similar manner as for a taxonomic group. For example, functional group can include a plurality of reference sequences corresponding to one or more genes of the functional group. Reference sequences of multiple bacteria for a same gene can correspond to a same functional group. Then, to determine the RAV, the number of sequence reads assigned to the functional group can be used to determine a proportion for the functional group. In exemplary embodiment, the functional group is a KEGG or COG group.
The use of a functional group, which may include a single gene, can help to identify situations where there is a small change (e.g., increase) in many taxonomic groups such that the individual changes are too small to be statistically significant. In such cases, the changes may all be for a same gene or set of genes of a same functional group, and thus the change for that functional group can be statistically significant, even though the changes for the taxonomic groups may not be statistically significant for a given sequence dataset. The reverse can be true of a taxonomic group being more predictive than a particular functional group, e.g., when a single taxonomic group includes many genes that have changed by a relatively small amount.
As an example, if 10 taxonomic groups increase by approximately 10%, the statistical power to discriminate between the two groups may be low when each taxonomic group is analyzed individually. But, if the increase is similar all for genes(s) of a shared functional group, then the increase would be 100%, or a doubling of the proportion for that ta.xonomic group. This large increase would have a much larger statistical power for discriminating between the two groups. Thus, the functional group can act to provide a sum of small changes for various taxonomic groups. And, small changes for various functional groups, which happen to all be on a same taxonomic group, can sum to provide high statistical power for that particular taxonomic group.
2. Exemplary Pipeline for Detecting and Analyzing Taxonomic Groups
Embodiments can provide a bioinformatics pipeline that taxonomically annotates the microorganisms present in a sample. The example clinical annotation pipeline can comprise the following procedures described herein.
In block 30, the samples can be identified and the sequence data can be loaded. For example, the pipeline can begin with demultiplexed fastq files (or other suitable files) that are the product of pair-end sequencing of amplicons (e.g., of the V4 region of the 16S gene). All samples can be identified for a given input sequencing file, and the corresponding fa.stq files can be obtained from the fastq repository server and loaded into the pipeline.
In block 31, the reads can be filtered. For example, a global quality filtering of reads in the fastq files can accept reads with a global Q-score >30. In one implementation, for each read, the per-position Q-scores are averaged, and if the average is equal or higher than 30, then the read is accepted, else the read is discarded, as is its paired read.
In block 32, primers can be identified and removed. In one embodiment, only forward reads that contain the forward primer and reverse reads that contain the reverse primer (allowing annealing of primers with up to 5 mismatches or other number of mismatches) are further considered. Primers and any sequences 5′ to them are removed from the reads. The 125 by (or other suitable number) towards the 3′ of the forward primer are considered from the forward reads, and only 124 bp (or other suitable number) towards the 3′ of the reverse primer are considered for the reverse reads. All processed forward reads that are <125bp and reverse reads that are <124bp are eliminated from further processing as are their paired reads.
In block 33, the forward and reverse reads can be written to files (e.g., FASTA files). For example, the forward and reverse reads that remained paired can be used to generate files that contain 125 bp from the forward read, concatenated to 124 bp from the reverse read (in the reverse complement direction).
In block 34, the sequence reads can be clustered, e.g., to identify chimeric sequences or determine a consensus sequence for a bacterium. For example, the sequences in the files can be subjected to clustering using the Swami algorithm [Mahé, F. et al. 2014] with a distance of 1. This treatment allows the generation of cluster composed of a central biological entity, surrounded by sequences which are 1 mutation away from the biological entity, which are less abundant and the result of the normal base calling error associated to high throughput sequencing. Singletons are removed from further analyses. In the remaining clusters, the most abundant sequence per cluster is then used as the representative and assigned the counts of all members in the cluster.
In block 35, chimeric sequences can be removed. For example, amplification of gene superfamilies can produce the formation of chimeric DNA sequences. These result from a partial PCR product from one member of the superfamily that anneals and extends over a different member of the superfamily in a subsequent cycle of PCR. In order to remove chimeric DNA sequences, some embodiments can use the VSEARCH chimera detection algorithm with the de nave option and standard parameters [Rognes, T. et al. 2016]. This algorithm uses abundance of PCR products to identify reference “real” sequences as those most abundant, and chimeric products as those less abundant and displaying local similarity to two or more of the reference sequences. All chimeric sequences can be removed from further analysis.
In block 36, taxonomy annotation can be assigned to sequences using sequence identity searches. To assign taxonomy to the sequences that have passed all filters above, some embodiments can perform identity searches against a database that contains bacterial strains (e.g., reference sequences) annotated to phylum, class, order, family, genus and species level, at least to a subsection of those taxonomic levels, or any other taxonomic levels. The most specific level of taxonomic annotation for a sequence can be kept, given that higher order taxonomy designations for a lower level taxonomy level can be inferred. The sequence identity search can be performed using the algorithm VSEARCH [Rognes, T. et al. 2016] with parameters (maxaccepts=0, maxrejects=0, id=1.) that allow an exhaustive exploration of the reference database used. Decreasing values of sequence identity can be used to assign sequences to different taxonomic groups: >97% sequence identity for assigning to a species, >95% sequence identity for assigning to a genus, >90% for assigning to family, >85% for assigning to order, >80% for assigning to class, and >77% for assigning to phylum.
In block 37, relative abundances of each taxa can be estimated and output to a database. For example, once all sequences have been used to identify identical sequences in the reference database, relative abundance per taxa can be determined by dividing the count of all sequences that are assigned to the same taxonomic group by the total number of reads that passed filters, e.g., were assigned. Results can be uploaded to database tables that are used as repositoty for the taxonomic annotation data.
3. Exemplary Pipeline for Detecting and Analyzing Functional Groups
For functional groups, the process can proceed as follows.
In block 40, sample OTUs (Operational Taxonomic Units) can be found. This may occur, e.g., after the sixth block described above in section V.B.2. After sample OTUs are found, sequences can be clustered, e.g., based on sequence identity (e.g., 97% sequence identity).
In block 41, a taxonomy can be assigned, e.g., by comparing OTUs with reference sequences of known taxonomy. The comparison can be based on sequence identity (e.g., 97%).
In block 42, taxonomic abundance can be adjusted for 16S copy number, or whatever genomic regions may be analyzed. Different species may have different number of copies of the 16S gene, so those possessing a higher number of copies will have more 16S material for PCR amplification at same number of cells than other species. Therefore, abundance can he normalized by adjusting the number of 16S copies.
In block 43, a pre-computed genomic lookup table can be used to relate taxonomy to functions, and amount of function. For example, a pre-computed genomic lookup table that shows the number of genes for important KEGO or COG functional categories per taxonomic group can be used to estimate the abundance of those functional categories based on the normalized 16S abundance data.
Upon identification of represented groups of microorganisms of the microhiome associated with a biological sample and/or identification of candidate functional aspects (e.g., functions associated with the microbiome components of the biological samples), generating features derived from compositional and/or functional aspects of the microbiome associated with the aggregate set of biological samples can be performed.
In one variation, generating features can include generating features derived from multilocus sequence typing (MLST), which can he performed experimentally at any stage in relation to implementation of the methods 100, 200, in order to identify markers useful for characterization in subsequent blocks of the method 100. Additionally or alternatively, generating features can include generating features that describe the presence or absence of certain taxonomic groups of microorganisms, and/or ratios between exhibited taxonomic groups of microorganisms. Additionally or alternatively, generating features can include generating features describing one or more of: quantities of represented taxonomic groups, networks of represented taxonomic groups, correlations in representation of different taxonomic groups, interactions between different taxonomic groups, products produced by different taxonomic groups, interactions between products produced by different taxonomic groups, ratios between dead and alive microorganisms (e.g., for different represented taxonomic groups, e.g., based. upon analysis of RNAs), phylogenetic distance (e.g., in terms of Kantorovich-Rubinstein distances, Wasserstein distances etc.), any other suitable taxonomic group-related feature(s), or any other suitable genetic or functional feature(s).
Additionally or alternatively, generating features can include generating features describing relative abundance of different microorganism groups, for instance, using a sparCC approach, using Genome Relative Abundance and Average size (GAAS) approach and/or using a genome Relative Abundance using Mixture Model theory (GRAMM) approach that uses sequence-similarity data to perform a maximum likelihood estimation of the relative abundance of one or more groups of microorganisms. Additionally or alternatively, generating features can include generating statistical measures of taxonomic variation, as derived from abundance metrics. Additionally or alternatively, generating features can include generating features derived from relative abundance factors (e.g., in relation to changes in abundance of a taxon, which affects abundance of other taxa). Additionally or alternatively, generating features can include generation of qualitative features describing presence of one or more taxonomic groups, in isolation and/or in combination. Additionally or alternatively, generating features can include generation of features related to genetic markers (e.g., representative 16S, 18S, and/or ITS sequences) characterizing microorganisms of the microbiome associated with a biological sample. Additionally or alternatively, generating features can include generation of features related to functional associations of specific genes and/or organisms having the specific genes. Additionally or alternatively, generating features can include generation of features related to pathogenicity of a taxon and/or products attributed to a taxon. Block S120 can, however, include generation of any other suitable feature(s) derived from sequencing and mapping of nucleic acids of a biological sample. For instance, the feature(s)can be combinatory (e.g., involving pairs, triplets), correlative (e.g., related to correlations between different features), and/or related to changes in features (i.e., temporal changes, changes across sample sites, spatial changes, etc.). Features can, however, be generated in any other suitable manner in Block S120.
4. Use of Supplementary Data
Block S130 recites: receiving a supplementary dataset, associated with at least a subset of the population of subjects, wherein the supplementary dataset is informative of characteristics associated with the disease or condition. The supplementary dataset can thus be informative of presence of the disease within the population of subjects. Block S130 functions to acquire additional data associated with one or more subjects of the set of subjects, which can be used to train and/or validate the characterization processes performed in block S140. In Block S130, the supplementary dataset can include survey-derived data, and can additionally or alternatively include any one or more of: contextual data derived from sensors, medical data (e.g., current and historical medical data associated with an oral health issue or health conditions associated with an oral health issue, dental X-ray data, results from a periodontal evaluation (e.g., ADA code D0120, or D0180), behavioral instrument data, data derived from a tool derived from the Diagnostic and. Statistical Manual of Mental Disorders, etc.), and any other suitable type of data.
In variations of Block S130 including reception of survey-derived data, the survey-derived data preferably provides physiological, demographic, and behavioral information in association with a subject. Physiological information can include information related to physiological features (e.g., height, weight, body mass index, body fat percent, body hair level, etc.). Demographic information can include information related to demographic features (e.g., gender, age, ethnicity, marital status, number of siblings, socioeconomic status, sexual orientation, etc.). Behavioral information can include information related to one or more of: health conditions (e.g., health and disease states), living situations (e.g., living alone, living with pets, living with a significant other, living with children, etc.), dietary habits (e.g., omnivorous, vegetarian, vegan, sugar consumption, acid consumption, etc.), behavioral tendencies (e.g., levels of physical activity, drug use, alcohol use, etc.), different levels of mobility (e.g., related to distance traveled within a given tune period), different levels of sexual activity (e.g., related to numbers of partners and sexual orientation), and any other suitable behavioral information. Survey-derived data can include quantitative data and/or qualitative data that can be converted to quantitative data (e.g., using scales of severity, mapping of qualitative responses to quantified scores, etc.).
In facilitating reception of survey-derived data, Block S130 can include providing one or more surveys to a subject of the population of subjects, or to an entity associated with a subject of the population of subjects. Surveys can be provided in person (e.g., in coordination with sample provision and/or reception from a subject), electronically (e.g., during account setup by a subject, at an application executing at an electronic device of a subject, at a web application accessible through an internet connection, etc. and/or in any other suitable manner.
Additionally or alternatively, portions of the supplementary dataset received in Block S130 can be derived from sensors associated with the subject(s) (e.g., sensors of wearable computing devices, sensors of mobile devices, biometric sensors associated with the user, etc.). As such, Block S130 can include receiving one or more of: physical activity- or physical action-related data (e.g., accelerometer and gyroscope data from a mobile device or wearable electronic device of a subject), environmental data (e.g., temperature data, elevation data, climate data, light parameter data, etc.), patient nutrition or diet-related data (e.g., data from food establishment check-ins, data from spectrophotometric analysis, etc.), biometric data (e.g., data recorded through sensors within the patient's mobile computing device, data recorded through a wearable or other peripheral device in communication with the patient's mobile computing device), location data (e.g., using GPS elements), and any other suitable data, Additionally or alternatively, portions of the supplementary dataset can be derived from medical record data and/or clinical data of the subject(s). As such, portions of the supplementary dataset can be derived from one or more electronic health records (EHRs) of the subject(s).
Additionally or alternatively, the supplementary dataset of Block S130 can include any other suitable diagnostic information e.g., clinical diagnosis information), which can be combined with analyses derived from features to support characterization of subjects in subsequent blocks of the method 100. For instance, information derived from a colonoscopy, biopsy, blood test, diagnostic imaging, survey-related information, and any other suitable test car be used to supplement Block S130.
5. Characterization of Oral Health Issues
Block S140 recites: transforming the supplementary dataset and features extracted from at least one of the microbiome composition dataset and the microbiome functional diversity dataset into a characterization model of the disease or condition. Block S140 functions to perform a characterization process for identifying features and/or feature combinations that can be used to characterize subjects or groups with an oral health issue based upon their microbiome composition and/or functional features. Additionally or alternatively, the characterization process can be used as a diagnostic tool that can characterize a subject (e.g., in terms of behavioral traits, in terms of medical conditions, in terms of demographic traits, etc.) based upon their microbiome composition and/or functional features, in relation to other health condition states, behavioral traits, medical conditions, demographic traits, and/or any other suitable traits. Such characterization can then be used to suggest or provide personalized therapies by way of the therapy model of Block S150.
In performing the characterization process, Block S140 can use computational methods (e.g., statistical methods, machine learning methods, artificial intelligence methods, bioinformatics methods, etc.) to characterize a subject as exhibiting features characteristic of a group of subjects with an oral health issue.
In one variation, characterization can be based upon features derived from a statistical analysis (e.g., an analysis of probability distributions) of similarities and/or differences between a first group of subjects exhibiting a target state (e.g., a health condition state) associated with the oral health issue, and a second group of subjects not exhibiting the target state (e.g., a “normal” state) associated with absence of an oral health issue, or the absence of a microbiome indicative of an oral health issue, or the absence of a microbiome indicative of a health and/or quality of life issue caused by an oral health issue. In implementing this variation, one or more of a Kolmogorov-Smirnov (KS) test, a permutation test, a Cramér-von Mises test, and any other statistical test (e.g., t-test, Welch's t-test, z-test, chi-squared test, test associated with distributions, etc.) can be used. In particular, one or more such statistical hypothesis tests can be used to assess a set of features having varying degrees of abundance in (or variations across) a first group of subjects exhibiting a target state (e.g., an adverse state) associated with the an oral health issue and a second group of subjects not exhibiting the target state (e.g., having a normal state) associated with oral health issue. In more detail, the set of features assessed can be constrained based upon percent abundance and/or any other suitable parameter pertaining to diversity in association with the first group of subjects and the second group of subjects, in order to increase or decrease confidence in the characterization. In a specific implementation of this example, a feature can be derived from a taxon of microorganism and/or presence of a functional feature that is abundant in a certain percentage of subjects of the first group and subjects of the second group, wherein a relative abundance of the taxon between the first group of subjects and the second group of subjects can be determined from one or more of a KS test or a Welch's t-test (e.g., a t-test with a log normal transformation), with an indication of significance (e.g., in terms of p- value). Thus, an output of Block S140 can comprise a normalized relative abundance value (e.g., 25% greater abundance of a taxon-derived feature and/or a functional feature in oral health issue subjects vs. control subjects) with an indication of significance (e.g., a p-value of 0.0013). Variations of feature generation can additionally or alternatively implement or be derived from functional features or metadata features (e.g., non-bacterial markers)
In variations and examples, characterization can use the relative abundance values (RAVs) for populations of subjects that have the disease (an oral health issue) and that do not have the disease (control population). If the distribution of RAVs of a particular sequence group for the disease population is statistically different than the distribution of RAVs for the control population, then the particular sequence group can be identified for including in a disease signature. Since the two populations have different distributions, the RAV for a new sample for a sequence group in the disease signature can be used to classify (e,g., determine a probability) of whether the sample does or does not have, or is indicative of, the disease. The classification can also be used to determine a treatment, as is described herein. A discrimination level can be used to identify sequence groups that have a high predictive value. Thus, embodiment can filter out taxonomic groups and/or functional groups that are not very accurate for providing a diagnosis.
Once RAVs of a sequence group have been determined for the control and disease populations, various statistical tests can be used to determine the statistical power of the sequence group for discriminating between disease (an oral health issue) and the absence of the disease (control). In one embodiment, the Kohnogorov-Smirnov (KS) test can be used to provide a probability value (p-value) that the two distributions are actually identical. The smaller the p-value the greater the probability to correctly identify which population a sample belongs. The larger the separation in the mean values between the two populations generally results in a smaller p-value (an example of a discrimination level). Other tests for comparing distributions can be used. The Welch's t-test presumes that the distributions are Gaussian, which is not necessarily true for a particular sequence group. The KS test, as it is a non- parametric test, is well suited for comparing distributions of taxa or functions for which the probability distributions are unknown.
The distribution of the RAVs for the control and disease populations can be analyzed to identify sequence groups with a large separation between the two distributions. The separation can be measured as a p-value (See example section). For example, the RAVs for the control population may have a distribution peaked at a first value with a certain width and decay for the distribution. And, the disease population can have another distribution that is peaked a second value that is statistically different than the first value. In such an instance, an abundance value of a control sample has a lower probability to be within the distribution of abundance values encountered for the disease samples. The larger the separation between the two distributions, the more accurate the discrimination is for determining whether a given sample belongs to the control population or the disease population. As is described herein, the distributions can be used to determine a probability for an RAV as being in the control population and determine a probability for the RAV being in the disease population, where sequence groups associated with the largest percentage difference between two means have the smallest p-value, signifying a greater separation between the two populations.
In performing the characterization process, Block S140 can additionally or alternatively transform input data from at least one of the microbiome composition datasets and/or microbiome functional diversity datasets into feature vectors that can be tested for efficacy in predicting characterizations of the population of subjects. Data from the supplementary dataset can be used to inform characterizations of the oral health issue, wherein the characterization process is trained with a training dataset of candidate features and candidate classifications to identify features and/or feature combinations that have high degrees (or low degrees) of predictive power in accurately predicting a classification. As such, refinement of the characterization process with the training dataset identifies feature sets (eg., of subject features, of combinations of features) having high correlation with an oral health issue or a health issue (e.g., symptom) associated with an oral health issue.
In some embodiments, feature vectors effective in predicting classifications of the characterization process can include features related to one or more of: microbiome diversity metrics (e.g., in relation to distribution across taxonomic groups, in relation to distribution across archa.eal, bacterial, viral, and/or eukaryotic groups), presence of taxonomic groups in one's microbiome, representation of specific genetic sequences (e.g., 16S sequences) in one's microbiome, relative abundance of taxonomic groups in one's microbiome, microbiome resilience metrics (e.g., in response to a perturbation determined from the supplementary dataset), abundance of genes that encode proteins or RNAs with given functions (enzymes, transporters, proteins from the immune system, hormones, interference RNAs, etc.) and any other suitable features derived from the microbiome composition dataset, the microbiome functional diversity dataset (e.g., COG-derived features, KEGG derived features, other functional features, etc.), and/or the supplementary dataset. Additionally, combinations of features can be used in a feature vector, wherein features can be grouped and/or weighted in providing a combined feature as part of a feature set. For example, one feature or feature set can include a weighted composite of the number of represented classes of bacteria in one's microbiome, presence of a specific genus of bacteria in one's microbiome, representation of a specific 16S sequence in one's microbiome, and relative abundance of a first phylum over a second phylum of bacteria. However, the feature vectors can additionally or alternatively be determined in any other suitable manner.
In examples of Block S140, assuming sequencing has occurred at a sufficient depth, one can quantify the number of reads for sequences indicative of the presence of a feature, thereby allowing one to set a value for an estimated amount of one of the criteria. The number of reads or other measures of amount of one of the features can be provided as an absolute or relative value. An example of an absolute value is the number of reads of 16S rRNA coding sequence reads that map to the genus of Lacnospira. Alternatively, relative amounts can be determined. An exemplary relative amount calculation is to determine the amount of 16S rRNA coding sequence reads for a particular bacterial taxon (e.g., genus , family, order, class, or phylum) relative to the total number of 16S rRNA coding sequence reads assigned to the bacterial domain. A value indicative of amount of a feature in the sample can then be compared to a cut-off value or a probability distribution in a disease signature for an oral health issue. For example, if the disease signature indicates that a relative amount of feature #1 of 50% or more of all features possible at that level indicates the likelihood of an oral health issue or a health or quality of life issue attributable to, indicative of, or caused by an oral health issue, then quantification of gene sequences associated with feature #1 less than 50% in a sample would indicate a higher likelihood of being from a healthy subject (or at least from a subject that does not have an oral health, or does not have a specific an oral health issue) and alternatively, quantification of gene sequences associated with feature #1 of more than 50% in a sample would indicate a higher likelihood of the disease.
In some cases, the taxonomic groups and/or functional groups can be referred to as features, or as sequence groups in the context of determining an amount of sequence reads corresponding to a particular group (feature). In some cases, scoring of a particular bacteria or genetic pathway can be determined according to a comparison of an abundance value to one or more reference (calibration) abundance values for known samples, e.g., where a detected abundance value less than a certain value is associated with the oral health issue in question and above the certain value is scored as associated with healthy, or vice versa depending on the particular criterion. The scoring for various bacteria or genetic pathways can be combined to provide a classification for a subject. Furthermore, in the examples, the comparison of an abundance value to one or more reference abundance values can include a comparison to a cutoff value determined from the one or more reference values. Such cutoff value(s) can be part of a decision tree or a clustering technique (where a cutoff value is used to determine which cluster the abundance value(s) belong) that are determined using the reference abundance values. The comparison can include intermediate determination of other values, (e.g., probability values) The comparison can also include a comparison of an abundance value to a probability distribution of the reference abundance values, and thus a comparison to probability values.
A disease signature can include more sequence groups than are used for a given subject. As an example, the disease signature can include 100 sequence groups, but only 60 of sequence groups may be detected in a sample, or detected above a threshold cutoff. The classification of the subject (including any probability for having or lacking a disease such as an oral health issue) can be determined based on the 60 sequence groups.
In relation to generation of the characterization model, the sequence groups with high discrimination levels (e.g., low p-values) for a given disease can be identified and used as part of a characterization model, e.g., which uses a disease signature to determine a probability of a subject having an oral health issue. The disease signature can include a set of sequence groups as well as discriminating criteria (e.g., cutoff values and/or probability distributions) used to provide a classification of the subject. The classification can be binary (e.g., disease or control) or have more classifications (e.g., probability values for having the disease of an oral health issue, or not having the disease). Which sequence groups of the disease signature that are used in making a classification be dependent on the specific sequence reads obtained, e.g., a sequence group would not be used if no sequence reads were assigned to that sequence group. In some embodiments, a separate characterization model can be determined for different populations, e.g., by geography where the subject is currently residing (e.g., country, region, or continent), the generic history of the subject (e.g., ethnicity), or other factors.
6. Selection of Sequence Groups, Discrimination Criteria for Sequence Groups, and Use of Sequence Groups
As shown in
In one implementation, a characterization process of Block S140 based upon statistical analyses can identify the sets of features that have the highest correlations with an oral health issue, for which one or more therapies would have a positive effect, based upon an algorithm trained and validated with a validation dataset derived from a subset of the population of subjects. In particular, an oral health issue in this first variation is characterized by an alteration of the microbiome that is predictive of the presence or absence of dental decay or predictive of the presence or absence of gingivitis.
In one variation, a set of features useful for diagnostics associated with oral health issues includes features derived from one or more of the taxa of TABLEs A or B (e.g., one or more of the family, order, class, and/or phylum of TABLE A, or the species of TABLE B) and/or one or more of the functional groups of TABLE B (e.g., one or more of the KEGG level 2 (KEGG L2) functional groups and/or one or more of the KEGG level 3 (KEGG L3) functional groups of TABLE B).
7. Therapy Models
In some embodiments, as noted above, outputs of the first method 100 can be used to generate diagnostics and/or provide therapeutic measures for an individual based upon an analysis of the individual's microbiome. As such, a second method 200 derived from at least one output of the first method 100 can include: receiving a biological sample from a subject S210, characterizing the subject with a form of an oral health issue based upon the characterization and the therapy model S230.
Block S210 recites: receiving a biological sample from the subject, which functions to facilitate generation of a microbiome composition dataset and/or a microbiome functional diversity dataset for the subject. As such, processing and analyzing the biological sample preferably facilitates generation of a microbiome composition dataset and/or a microbiome functional diversity dataset for the subject, which can be used to provide inputs that can be used to characterize the individual in relation to diagnosis of the oral health issue, as in Block S220. Receiving a biological sample from the subject is preferably performed in a manner similar to that of one of the embodiments, variations, and/or examples of sample reception described in relation to Block S110 above. As such, reception and processing of the biological sample in Block S210 can be performed for the subject using similar processes as those for receiving and processing biological samples used to generate the characterization(s) and/or the therapy provision model of the first method 100, in order to provide consistency of process. However, biological sample reception and processing in Block S210 can alternatively be performed in any other suitable manner.
Block S220 recites: characterizing the subject characterizing the subject with a form of a disease or condition based upon processing a microbiome dataset derived from the biological sample. Block S220 functions to extract features from microbiome-derived data of the subject, and use the features to positively or negatively characterize the individual as having a form of the oral health issue. Characterizing the subject in Block S220 thus preferably includes identifying features and/or combinations of features associated with the microbiome composition and/or functional features of the microbiome of the subject, and comparing such features with features characteristic of subjects with the oral health issue. Block S220 can further include generation of and/or output of a confidence metric associated with the characterization for the individual. For instance, a confidence metric can be derived from the number of features used to generate the classification, relative weights or rankings of features used to generate the characterization, measures of bias in the models used in Block S140 above, and/or any other suitable parameter associated with aspects of the characterization operation of Block S140.
In some variations, features extracted from the microbiome dataset can be supplemented with survey-derived and/or medical history-derived features from the individual, which can be used to further refine the characterization operation(s) of Block S220. However, the microbiome composition dataset and/or the microbiome functional diversity dataset of the individual can additionally or alternatively be used in any other suitable manner to enhance the first method 100 and/or the second method 200.
Block S230 recites: promoting a therapy to the subject with the disease or condition based upon the characterization and the therapy model. Block S230 functions to recommend or provide a personalized therapeutic measure to the subject, in order to shift the microbiome composition of the individual toward a desired equilibrium state. As such, Block S230 can include correcting the oral health issue, or otherwise positively affecting the user's health in relation to the oral health issue. Block S230 can thus include promoting one or more therapeutic measures to the subject based upon their characterization in relation to the oral health issue, as described herein, wherein the therapy is configured to modulate taxonomic makeup of the subject's microbiome and/or modulate functional feature aspects of the subject in a desired manner toward a “normal” or “control” state in relation to the characterizations described above.
In Block S230, providing the therapeutic measure to the subject can include recommendation of available therapeutic measures configured to modulate microbiome composition of the subject toward a desired state having a microbiome that is not indicative of (e.g., altered by) an oral health issue). Additionally or alternatively, Block S230 can include provision of customized therapy to the subject according to their characterization (e.g., in relation to a specific type of an oral health issue, such as gingivitis or dental decay). In variations, therapeutic measures for adjusting a microbiome composition of the subject, in order to improve a state of the oral health issue can include one or more of: probiotics, prebiotics, bacteriophage-based therapies, consumables, suggested activities, topical therapies, adjustments to hygienic product usage, adjustments to diet, adjustments to sleep behavior, living arrangement, adjustments to level of sexual activity, nutritional supplements, medications, antibiotics, and any other suitable therapeutic measure. Therapy provision in Block S230 can include provision of notifications by way of an electronic device, through an entity associated with the individual, and/or in any other suitable manner.
In more detail, therapy provision in Block S230 can include provision of notifications to the subject regarding recommended therapeutic measures and/or other courses of action, in relation to health-related goals, as shown in
Furthermore, in an extension of Block S230, monitoring of the subject during the course of a therapeutic regimen (e.g., by receiving and analyzing biological samples from the subject throughout therapy, by receiving survey-derived data from the subject throughout therapy) can be used to generate a therapy-effectiveness model for each recommended therapeutic measure provided according to the model generated in Block S150.
As shown in
For instance, in relation to the variations of oral health issues described herein, therapies (e.g., probiotic therapies, bacteriophage-based therapies, prebiotic therapies, etc.) can be configured to downregulate and/or upregulate microorganism populations or subpopulations (and/or functions thereof) associated with features characteristic of the oral health issue.
In one such variation, the Block S150 can include one or more of the following steps: obtaining a sample from the subject; purifying nucleic acids (e.g., DNA) from the sample; deep sequencing nucleic acids from the sample so as to determine the amount of one or more of the features of TABLEs A or B; and comparing the resulting amount of each feature to one or more reference amounts of the one or more of the features listed in one or more of TABLEs A or B as occurs in an average individual having an oral health issue or an individual not having the oral health issue or both. The compilation of features can sometimes be referred to as a “disease signature” for a specific condition related to an oral health issue. The disease signature can act as a characterization model, and may include probability distributions for control population (no oral health issue) or disease populations having the condition or both. The disease signature can include one or more of the features (e.g., bacterial taxa or genetic pathways) listed and can optionally include criteria determined from abundance values of the control and/or disease populations. Example criteria can include cutoff or probability values for amounts of those features associated with average control or disease (e.g., dental decay or gingivitis) individuals.
In a specific example of probiotic therapies, as shown in
In variations, the therapy model is preferably based upon data from a large population of subjects, which can comprise the population of subjects from which the microbiome-related datasets are derived in Block S110, wherein microbiome composition and/or functional features or states of health, prior exposure to and post exposure to a variety of therapeutic measures, are well characterized. Such data can be used to train and validate the therapy provision model, in identifying therapeutic measures that provide desired outcomes for subjects based upon different microbiorne characterizations. In variations, support vector machines, as a supervised machine learning algorithm, can be used to generate the therapy provision model. However, any other suitable machine learning algorithm described above can facilitate generation of the therapy provision model.
While some methods of statistical analyses and machine learning are described in relation to performance of the Blocks above, variations of the method 100, or any one of
Additionally or alternatively, the therapy model can be derived in relation to identification of a “normal” or baseline microbiome composition and/or functional features, as assessed from subjects of a population of subjects who are identified to be in good health. Upon identification of a subset of subjects of the population of subjects who are characterized to be in good health (e.g., characterized as not having an altered microbiome caused by, or indicative of, an oral health issue, e.g., using features of the characterization process), therapies that modulate microbiome compositions and/or functional features toward those of subjects in good health can be generated in Block S150. Block S150 can thus include identification of one or more baseline microbiome compositions and/or functional features (e.g., one baseline microbiome for each of a set of demographics), and potential therapy formulations and therapy regimens that can shift microbiomes of subjects who are in a state of dysbiosis toward one of the identified baseline microbiome compositions and/or functional features. The therapy model can, however, be generated and/or refined in any other suitable manner.
Microorganism compositions associated with probiotic therapies associated with the therapy model preferably include microorganisms that are culturable (e.g., able to be expanded to provide a scalable therapy) and non-lethal (e.g., non-lethal in their desired therapeutic dosages). Furthermore, microorganism compositions can comprise a single type of microorganism that has an acute or moderated effect upon a subject's microbiome. Additionally or alternatively, microorganism compositions can comprise balanced combinations of multiple types of microorganisms that are configured to cooperate with each other in driving a subject's microbiome toward a desired state. For instance, a combination of multiple types of bacteria in a probiotic therapy can comprise a first bacteria type that generates products that are used by a second bacteria type that has a strong effect in positively affecting a subject's microbiome. Additionally or alternatively, a combination of multiple types of bacteria in a probiotic therapy can comprise several bacteria types that produce proteins with the same functions that positively affect a subject's microbiome.
In examples of probiotic therapies, probiotic compositions can comprise components of one or more of the identified taxa of microorganisms (e.g., as described in TABLE A) provided at dosages of 1 million to 10 billion CFUs, as determined from a therapy model that predicts positive adjustment of a subject's microbiome in response to the therapy. Additionally or alternatively, the therapy can comprise dosages of proteins resulting from functional presence in the microbiome compositions of subjects without the oral health issue. In the examples, a subject can be instructed to ingest capsules comprising the probiotic formulation according to a regimen tailored to one or more of his/her: physiology (e.g., body mass index, weight, height), demographics (e.g., gender, age), severity of dysbiosis, sensitivity to medications, and any other suitable factor.
Furthermore, probiotic compositions of probiotic-based therapies can be naturally or synthetically derived. For instance, in one application, a probiotic composition can be naturally derived from fecal matter or other biological matter (e.g., of one or more subjects having a baseline rnicrobiorne composition and/or functional features, as identified using the characterization process and the therapy model). Additionally or alternatively, probiotic compositions can be synthetically derived (e.g., derived using a benchtop method) based upon a baseline microbiome composition and/or functional features, as identified using the characterization process and the therapy model. In one embodiment, the probiotic composition is or is derived from the subject's own fecal matter that has been stored or “banked” from a period during which the subject is in a healthy state for use when the microbiome is imbalanced (e.g., due to antibiotic usage, or due to an oral health issue).
In variations, microorganism agents that can be used in probiotic therapies can include one or more of: yeast (e.g., Saccharomyces boulardii), gram-negative bacteria (e.g., E. coli Nissle, Akkermansia muciniphila, Prevotella bryantii, etc.), gram-positive bacteria (e.g., Bifidobacterium animalis (including subspecies lactis), Bifidobacterium longum (including subspecies infantis), Bifidobacterium bifidum, Bifidobacterium pseudolongum, Bifidobacterium thermophilum, Bifidobacterium breve, Lactobacillus rhamnosus, Lactobacillus acidophilus, Lactobacillus casei, Lactobacillus heiveticus, Lactobacillus plantarum, Lactobacillus fermentum, Lactobacillus salivarius, Lactobacillus deibrueckii (including subspecies bulgaricus), Lactobacillus johnsonii, Lactobacillus reuteri, Lactobacillus gasseri, Lactobacillus brevis (including subspecies coagulans), Bacillus cereus, Bacillus subtilis (including var. Natto), Bacillus polyfermenticus, Bacillus clausii, Bacillus licheniformis, Bacillus coagulans, Bacillus pumilus, Faecalibacterium prausnitzii, Streptococcus thermophilus, Brevibacillus brevis, Lactococcus lactis, Leuconsotoc mesenteroides, Enterococcus faecium, Enterococcus faecalis, Enterococcus durans, Clostridium butyricum, Sporolactobacillus inulinus, Sporolactobacillus vineae, Pediococcus acidilactici, Pediococcus pentosaceus, etc.), and any other suitable type of microorganism agent.
Additionally or alternatively, therapies promoted by the therapy model of Block S150 can include one or more of: consumables (e.g., food items, beverage items, nutritional supplements), suggested activities (e.g., exercise regimens, adjustments to alcohol consumption, adjustments to cigarette usage, adjustments to drug usage), topical therapies (e.g., lotions, ointments, antiseptics, etc.), adjustments to hygienic product usage (e.g., use of shampoo products, use of conditioner products, use of soaps, use of makeup products, etc.), adjustments to diet (e.g., sugar consumption, fat consumption, salt consumption, acid consumption, etc.), adjustments to sleep behavior, living arrangement adjustments (e.g., adjustments to living with pets, adjustments to living with plants in one's home environment, adjustments to light and temperature in one's home environment, etc.), nutritional supplements (e.g., vitamins, minerals, fiber, fatty acids, amino acids, prebiotics, probiotics, etc.), medications, antibiotics, and any other suitable therapeutic measure. Among the prebiotics suitable for treatment, as either part of any food or as supplement, are included the following components: 1,4-dihydroxy-2-naphthoic acid (DHNA), Inulin, trans-Galactooligosaccharides (GOS), Lactulose, Mannan oligosaccharides (MOS), Fructooligosaccharides (FOS), Neoagaro-oligosaccharides (NAOS), Pyrodextrins, Xylo-oligosaccharides (XOS), Isomalto-oligosaccharides (IMOS), Amylose-resistant starch, Soybean oligosaccharides (SBOS), Lactitol, Lactosucrose (LS), Isomaltulose (including Palatinose), Arabinoxylooligosaccharides (AXOS), Raffinose oligosaccharides (RFO), Arabinoxylans (AX), Polyphenols or any other compound capable of changing the microbiota composition with a desirable effect.
Additionally or alternatively, therapies promoted by the therapy model of Block S150 can include one or more of: different forms of therapy having different therapy orientations (e.g., motivational, increase energy level, reduce weight gain, improve diet, psychoeducational, cognitive behavioral, biological, physical, mindfulness-related, relaxation-related, dialectical behavioral, acceptance-related, commitment-related, etc.) configured to address a variety of factors contributing to an adverse states due to a microbiome that is altered by an oral health issue or a microbiome that is caused by or indicative of an oral health issue; weight management interventions (e.g., to prevent adverse weight-related (e.g., weight gain or loss) side effects due to dental decay or gingivitis, or a therapy to prevent, mitigate, or reduce the frequency or severity of dental decay or gingivitis); gingival graft; dental restoration; application of dental sealants; physical therapy; rehabilitation measures; and any other suitable therapeutic measure.
The first method 100 can, however, include any other suitable blocks or steps configured to facilitate reception of biological samples from individuals, processing of biological samples from individuals, analyzing data derived from biological samples, and generating models that can be used to provide customized diagnostics and/or therapeutics according to specific microbiome compositions of individuals.
The methods 100, 200 and/or system of the embodiments can be embodied and/or implemented at least in part as a machine configured to receive a computer-readable medium storing computer-readable instructions. The instructions can be executed by computer-executable components integrated with the application, a.pplet, host, server, network, website, communication service, communication interface, hardware/firmware/software elements of a patient computer or mobile device, or any suitable combination thereof. Other systems and methods of the embodiments can be embodied and/or implemented at least in part as a machine configured to receive a computer-readable medium storing computer-readable instructions. The instructions can be executed by computer-executable components integrated with apparatuses and networks of the type described above. The computer-readable medium can be stored on any suitable computer readable media such as RAMs, ROMs, flash memory, EEPROMs, optical devices (CD or DVD), hard drives, floppy drives, or any suitable device. The computer-executable component can be a processor, though any suitable dedicated hardware device can (alternatively or additionally) execute the instructions.
The FIGS illustrate the architecture, functionality and operation of possible implementations of systems, methods and computer program products according to preferred embodiments, example configurations, and variations thereof. In this regard, each block in the flowchart or block diagrams may represent a module, segment, step, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block can occur out of the order noted in the Figs. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
Example for Dental Decay
Some examples of sequence groups, discriminating levels, coverage percentages, and discriminating criteria are provided in TABLE A.
TABLE A shows data for dental decay. The data was obtained from 316 subjects in the condition population and 1107 subjects in the control population. TABLE A shows taxonomic groups for Family, Order, Class, and Phylum all in the first column of TABLE A. Each of the rows containing data corresponds to a different sequence group. For example, Pasteurellaceae corresponds to a sequence group in the Family level of the taxonomic hierarchy.
TABLE A shows a single sequence group for the Family level. A level can have many sequence groups. The number “712” after “Pasteurellaceae” is the NCBI taxonomy ID for that taxonomic group. The IDs correspond to those at www.ncbi.nlm.nih.gov/Taxonomy/Browser/wwwtax.cgi?id=200643. The p-values are determined via either the Kolmogorov-Smirnov test, or the Welch's t-test.
Sequence groups having a p-value less than 0.01 are shown in the second column. Other sequence groups may exist, but likely would not be selected for inclusion into a disease signature. The third column (“# disease subjects detected”) shows the number of samples tested that had the condition of dental decay and where the sample exhibited bacteria in the sequence group. The fourth column (“# control subjects detected”) shows the number of samples tested that did not have the disease (control) and where the sample exhibited bacteria in the sequence group. The coverage percentage of the sequence group can be determined from the values in the third and fourth columns.
The fifth column shows the mean percentage for the abundance for the subjects having the disease and where the sample exhibited bacteria in the sequence group. The sixth column shows the mean percentage for the abundance for the subjects not having the disease and where the sample exhibited bacteria in the sequence group. As one can see, the sequence groups with the largest percentage difference between the two means have the smallest p-value, signifying a greater separation between the two populations.
A set of sequence groups (taxonomic and/or functional) can be selected from TABLE A for forming a disease signature that can be used to classify a sample regarding a presence or absence of a microbiome indicative of a dental decay issue. For example, all four taxonomic sequence groups can be selected. The sequence groups for the disease signature can be selected to optimize accuracy for discriminating between the two groups and coverage of the population such that a likelihood of being able to provide a classification is higher (e.g., if a sequence group is not present then that sequence group cannot be used to determine the classification). The total coverage can dependent on the individual coverage percentages and based on the overlap in the coverages among the sequence groups, as described above.
A. Example for Gingivitis
Some examples of sequence groups, discriminating levels, coverage percentages, and discriminating criteria are provided in TABLE B.
TABLE B shows data for gingivitis. The data was obtained from two sub-populations (subset A and subset B). In subset A 130 subjects are in the condition population and 1110 subjects are in the control population. In subset B 212 subjects are in the condition population and 2067 subjects are in the control population. TABLE B shows the taxonomic group for Species, and shows functional groups for 1 KEGG L2 functional group, and 22 KEGG L3 functional groups all in the first column of TABLE B. As mentioned above, the functional groups correspond to one or more genes with the function. Each of the rows containing data corresponds to a different sequence group. For example, Cardiobacterium hominis corresponds to a sequence group in the Species level of the taxonomic hierarchy.
TABLE B shows a single sequence group for the Species level. A level can have many sequence groups. The number “2718” after “Cardiobacterium hominis” is the NCBI taxonomy ID for that taxonomic group. The IDs correspond to those at www.ncbi.nlm.nih.gov/Taxonomy/Browser/wwwtax.cgi?id=200643. The p-values are determined via either the Kolmogorov-Smirnov test, or the Welch's t-test.
Sequence groups having a p-value less than 0.01 are shown in the second column. Other sequence groups may exist, but likely would not be selected for inclusion into a disease signature. The third column (“# disease subjects detected”) shows the number of samples tested that had the condition of gingivitis and where the sample exhibited bacteria in the sequence group. The fourth column (“# control subjects detected”) shows the number of samples tested that did not have the disease (control) and where the sample exhibited bacteria in the sequence group. The coverage percentage of the sequence group can be determined from the values in the third and fourth columns.
The fifth column shows the mean percentage for the abundance for the subjects having the disease and where the sample exhibited bacteria in the sequence group. The sixth column shows the mean percentage for the abundance for the subjects not having the disease and where the sample exhibited bacteria in the sequence group. As one can see, the sequence groups with the largest percentage difference between the two means have the smallest p-value, signifying a greater separation between the two populations.
A set of sequence groups (taxonomic and/or functional) can be selected from TABLE B for forming a disease signature that can be used to classify a sample regarding a presence or absence of a microbiome indicative of a gingivitis issue. For example, 6 sequence groups can be selected, as may occur if the Cardiobacterium hominis Species taxonomic group and 5 KEGG L3 functional groups are selected. The sequence groups for the disease signature can be selected to optimize accuracy for discriminating between the two groups and coverage of the population such that a likelihood of being able to provide a classification is higher (e.g., if a sequence group is not present then that sequence group cannot be used to determine the classification). The total coverage can dependent on the individual coverage percentages and based on the overlap in the coverages among the sequence groups, as described above.
Although the foregoing invention has been described in some detail by way of illustration and example for purposes of clarity of understanding, one of skill in the art will appreciate that certain changes and modifications may be practiced within the scope of the appended claims. in addition, each reference provided herein is incorporated by reference in its entirety to the same extent as if each reference was individually incorporated by reference. Where a conflict exists between the instant application and a reference provided herein, the instant application shall dominate.
The present patent application claims benefit of priority to U.S. Provisional Application No. 62/215,924, filed Sep. 9, 2015, and U.S. Provisional Application No. 62/215,909, filed Sep. 9, 2015, the disclosures of each which are incorporated herein in the entirety and for all purposes.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/US16/51175 | 9/9/2016 | WO | 00 |
| Number | Date | Country | |
|---|---|---|---|
| 62215924 | Sep 2015 | US | |
| 62215909 | Sep 2015 | US |
