Patent Application
20220236274
The present invention relates to a novel method for the early diagnosis of pancreatic adenocarcinoma, comprising the use of gold nanoparticles and a subsequent image analysis, a method of screening or monitoring pathological and non-pathological patients to identify pathological patients at risk of pancreatic adenocarcinoma, comprising the use of gold nanoparticles and a subsequent image analysis, a kit comprising reactants for the early diagnosis of pancreatic adenocarcinoma.
Pancreatic adenocarcinoma is a cancer burdened by fatal prognosis and with a progressively increasing incidence. The time pattern of the incidence of this neoplasia, without considering population age variations, is on the rise both in female and male individuals (+1% and +1.3%/year, respectively). A further increase of new cases of the disease is expected in the next years, from 12.180 of 2020 to 13.928 of 2030. The only possibility of treatment, with a 5-year survival of 15%-20%, is offered by the surgical removal of the disease. Surgical resection, however, unfortunately to date is implementable in only 20% of patients. At diagnosis, in fact, the surgical approach is precluded to about 80-95% of the subjects, due to the presence of local invasion of the vascular structures, or to concomitant presence of distant metastases. Pancreatic adenocarcinoma exhibits insidious behavior, frequently asymptomatic or with entirely non-specific symptoms in the early stages and demonstrates a rapid progression to more advanced stages.
However, it has been demonstrated that about 15 years are necessary for the sequence of genome alterations intervening between the tumor origin and its metastatization to be completed (Yachida S, Jones S, Bozic I, et al. Distant metastasis occurs late during the genetic evolution of pancreatic cancer. Nature. 2010; 467(7319):1114-7). This type of clinical history, along with biological features, make the disease highly lethal. From the foregoing, there emerges the absolute need to be able to identify the disease in an early stage transpires, yet unfortunately to date there are no methods to be used for the early diagnosis that be scarcely invasive and having high sensitivity and specificity. The lack of sensitive and specific biomarkers induced numerous experts to search for a molecular signature typing the tumor by proteomic profiling approach in mass spectrometry.
With the use of nanomaterials in biology and medicine, it has been discovered that nanoparticles, contacted with biological fluids, adsorb on their surfaces (depending on their size, shape and material of which they are made up), a protein layer, developing what is described as “protein crown” (or “protein corona”).
It has recently been reported in the literature that nanoparticles (i.e., particles with at least one characteristic size ranging from 1 to 100 nm), after interaction with human plasma, coat themselves with a protein layer referred to as “protein crown”. Crown (Corona) formation is a complex process in which a delicate balance between electrostatic, hydrophobic and van der Waals interactions regulates its final protein composition. Recent research demonstrate that protein crown composition mainly depends on three factors: i) the chemico-physical properties of the nanoparticles (e.g., size, charge, material, surface chemistry); ii) the protein composition of the biological fluid; iii) environmental factors (e.g., temperature and exposure time). It has been demonstrated that the composition of the “protein crown” does not mirror slavishly human proteome composition but depends on the affinity of each specific protein for the particle surface. It follows that proteins scarcely abundant in human blood, yet highly affine to the nanoparticle surface, can be far more enriched in the protein crown compared to proteins highly abundant in human blood (like, e.g., serum albumin), but scarcely affine to the nanoparticle surface. Recently, Harvard Medical School researchers have demonstrated that the protein crown (corona) forming on the nanoparticle surface after interaction with human plasma is “personalized” (Hajipour, Mohammad Javad, et al. “Personalized disease-specific protein corona influences the therapeutic impact of graphene oxide.” Nanoscale 7.19 (2015): 8978-8994). This means that protein crown composition varies subjectwise as a result of individual variation in proteome composition. In fact, the analysis of changes in the proteome allows to identify proteins associated with the presence of a pathological condition, that may act as early-phase diagnostic markers, prognosis indicators or predictors of response to a treatment. Recently, many studies on the protein crown at physiological pH have become available, while little is still known on quali/quantitative crown modification under acidic conditions such as those encountered in inflamed tissues, tumor masses, lysosomes.
The non-patent document Colapicchioni, et al “Personalized liposome-protein corona in the blood of breast, gastric and pancreatic cancer patients” International Journal of Biochemistry and Cell Biology 2016, 75, 180-187 discloses that protein coronas isolated from the plasma of patients suffering from breast tumor, stomach tumor and pancreatic tumor are statistically different from each other. However, no image processing was carried out, nor any test able to predict a sample's belonging to the group of healthy patients or of the oncological patients was developed.
The non-patent document Caputo et al “A protein corona-enabled blood test for early cancer detection” Nanoscale, 2017, 9, 349-354 discloses the possibility to use the crown forming around anionic liposomes comprised of the dioleoyl-phosphatidyl-glycerol (DOPG) lipid to discriminate healthy subjects from patients suffering from pancreatic adenocarcinoma. Proof of feasibility as to the possibility of developing a blood test was provided.
In the state of the art, instead, a possible use of the protein crown depositing on nanoparticles for the early diagnosis of pancreatic adenocarcinoma is not known.
Current attempts at developing screening tests for the early diagnosis of pancreatic adenocarcinoma have mainly focused on serum biomarkers. Ca 19.9 is the only marker approved by the Food and Drug Administration in clinical practice for pancreatic adenocarcinoma. However, CA19-9 use as screening tool for pancreatic cancer in the general population in unacceptable owing to its low sensitivity (median 79%, with a range of 70-90%) and specificity (median 82%, with a range of 68-91%).
Precisely with the aim of increasing the diagnostic accuracy of Ca 19.9, in the last few years various biomarker panels have been identified, which by combining various proteins thereamong, with or without Ca 19.9, proved useful in discriminating the subjects suffering from pancreatic adenocarcinoma compared to healthy controls or subjects suffering from other pathologies, such as chronic pancreatitis. Some of these panels proved also able to distinguish tumors in their initial stage, as compared to more advanced ones. Unfortunately, to date these methodologies are not applicable on a large scale in daily clinical practice, as they are the result of highly complex and costly technologies and require highly specialized personnel.
In comparison with serum markers other modes were taken into consideration, such as marker collection and study on pancreatic juice. For this purpose, a more costly and invasive procedure is needed, endoscopic retrograde cholangiopancreatography (ERCP) or endoscopic ultrasound-guided fine-needle aspiration (EUS-FNA). These techniques, and in particular ERCP, are, among other things, burdened by a non-negligible rate of iatrogenic acute pancreatitis. Some biomarkers have also been identified by analyzing ductal brushing and cytology collections obtained by ERCP or FNA. However, in these cases as well, a sensitivity sufficiently high to justify the use of techniques so invasive and costly to diagnose early-stage pancreatic cancer was not reached.
Various imaging methods are used to identify neoplasias in patients that are symptomatic or with strongly suspected pancreatic neoplasia. The main imaging modes for pancreatic tumor detection are abdominal echography, echoendoscopy, CT, magnetic resonance, ERCP and positron-emission tomography (PET). Echography is non-invasive, easily available and does not expose the patient to ionizing radiations. However, owing to the position of the pancreas in the retroperitoneum, its accuracy is of <70%. Instead, transgastrically-performed echoendoscopy has a >95% sensitivity. Unfortunately, as with all echographic examinations, both are affected by the operator-related variability issue. CT with contrast media (cm) administration has a high sensitivity (90%) and specificity (99%), with a lower interobserver variability compared to echography. However, CT exposes patients to ionizing radiations and, due to the requirement of intravenous contrast, is not ideal for use in all patients, especially those with kidney failure or allergies. Magnetic resonance has a sensitivity and a specificity similar to CT, yet it has longer acquisition times and requires greater collaboration by patients, who have to keep still in order to obtain an accurate image. In general, high costs, invasiveness and the high rate of false positives represent the main drawbacks of the use of imaging techniques in screening programs of most tumors, and above all of pancreatic tumor.
Though to date there is an incomplete understanding of risk factors for pancreatic cancer, some groups have since years been identified as high-risk on the basis of clinical and genetic characteristics, and they might represent the target of application of the proposed technology. Clinical risk factors include age, obesity, smoking, diabetes and chronic pancreatitis. Pancreatic ductal adenocarcinoma appears in >95% of cases after 45 years of age, obese subjects have an up to 1.8-fold higher risk of developing the disease, as well as smokers and those who have quit smoking since less than 5 years, and those who have documented familiarity. A 1.76-fold higher risk is faced by diabetic patients, and new-onset diabetes can represent an early indicator of pancreatic neoplasia. In patients suffering from chronic pancreatitis the risk of tumor is more than doubled. Instead, as to genetic risk factors, BRCA2 mutations are associated with an increase of the risk of cancer of at least 3,5- and up to 10-fold, whereas a possible link with BRCA1 mutations has been observed. Other genetic risk factors include Peutz-Jeghers syndrome (STK11/LKB1 mutations) with a 132-fold risk, familial atypical multiple mole melanoma syndrome (CDKN2A mutations) with 13-22-fold increase of the risk, familial adenomatous polyposis (FAP) with a risk increased 4-fold, Lynch syndrome, with a risk increased 8.6-fold, and hereditary pancreatitis (PRSS1 mutations) with a 53-fold risk.
The Authors of the present invention have discovered that by analyzing the protein crown resulting from contacting gold nanoparticles with blood samples of patients suffering from adenocarcinoma it was possible to identify anomalies in the overall protein expression by low-cost techniques, such as, e.g., polyacrylamide gel electrophoresis (PAGE) of proteins by using sodium-dodecyl-sulfate (SDS). SDS-PAGE is a low-cost, fast experimental technique, widespread in clinical laboratories. The low resolution of electrophoresis, though not allowing to identify individual tumor biomarkers (as is done with mass spectrometry), enabled to assess “overall variations” in the protein profile even without a priori hypotheses on possible alterations in question, and in the detection of a profile of multiple proteins whose expression is modified overall in connection with the presence of the pathology.
In short, the present invention is based on the difference detected in the overall composition of the protein crown formed around gold nanoparticles of size equal to 100 nm after interaction with human plasma from healthy subjects and patients suffering from pancreatic adenocarcinoma. Crown overall composition was characterized by SDS-PAGE experiments. The study whose results are reported in the present patent application was financed by AIRC (Italian Association for Cancer Research) (IG 2017 ID. 20327; Principal Investigator: Prof. Giulio Caracciolo) and approved beforehand by the Ethics Committee of the Universitá Campus Bio-Medico of Rome (Protocol number: 10.3(12).18; Amendment 3 of Sep. 18, 2018).
Compared to what described in Caputo et al 2017, the invention differs in: i) the use of gold nanoparticles in lieu of liposomes; ii) the gel image processing procedure (absent in the cited work); iii) the methodology of one-dimensional protein profile processing and input variable identification. The use of gold nanoparticles in lieu of liposomes is essential for the clinical application of the blood test. In fact, colloidal dispersions of gold nanoparticles are commercial, have a low cost, are ready for use and stable over time. Conversely, liposomes need a complex preparation procedure from powdered lipid species, required a dedicated instrumentation (balance, rotavapor, ultrasound extrusor/sonicator) and are stable over a short period of time if stored at 4° C.
No indication is provided, neither in the prior art nor in the cited document, for image processing, experimental data treatment and statistical techniques of experimental data treatment so as to obtain relevant results in the early diagnosis of pancreatic adenocarcinoma.
Therefore, the present invention relates to a method to assist in the early diagnosis of pancreatic adenocarcinoma which allows the identification of subjects in an early state of disease or at risk of pancreatic adenocarcinoma for whom it is necessary or appropriate to carry out the complex examinations conventionally used today (herein referred to as second-level) which normally are carried out in an often belated manner, no tests being available, to date, which allow a screening to identify in populations potentially at risk of disease, potentially diseased patients (in the present description also referred to as potentially pathological) and patients for whom the actual risk be validated by laboratory tests and not merely by the patient's medical and family history. As to date no diagnostic tests are available which allow examinations for the carrying out of an early screening of the risk or of the potential presence of the pancreatic tumor, the test subject-matter of patent would therefore represent a first-level test, for a better identification of patients already potentially diseased, or in which the risk of disease be effectively validated by a laboratory examination. The examinations currently carried out, when the general pathological conditions of the patient indicate a presumable state of disease (and therefore normally when the disease is actually present and at an advanced stage) are defined herein as second-level examinations, and are represented by CT, MRI, ERCP and echoendoscopy. The method of the invention is based on a classifying of a plasma sample of these subjects according to two or more levels of risk in which the classification of said sample in at least one of said levels of risk indicates the need or opportunity to carry out second-level examinations, comprising the following steps:
Object of the invention are also a computer program comprising code adapted to carry out steps e) to h) as defined in the present description and in the claims when executed on a computer, a storage medium comprising said program, and a kit to assist in the early diagnosis of pancreatic adenocarcinoma that allows the identification of subjects in an early state of disease or at risk of pancreatic adenocarcinoma for whom it is necessary or appropriate to carry out second-level examinations, comprising one or more aliquots of one or more reactants adapted to carry out the steps a) to d) and optionally b′) and a storage medium as defined above.
In the present description the term protein crown (corona) has the meaning commonly used in the scientific literature, and it denotes the layer of proteins adsorbed on the surface of the nanoparticles following interaction with biological fluids (e.g., blood and plasma).
For the purposes of the invention and of the present description, therefore, the first-level examination is represented by the examination carried out with the method of the present invention, whereas the second-level examinations are represented by those conventionally carried out to date, which are at least one among abdomen CT, abdomen MRI, echoendoscopy, ERCP (endoscopic retrograde cholangiopancreatography).
Notably, said second-level examinations, being complex and costly for the patient or the national health system, are normally carried out when the patient's general health conditions show concrete symptoms of the disease, and are therefore normally carried out at a late stage of the disease, which, as indicated in the discussion of the state of the art, usually leads to the patient's death.
For the purposes of the present description, by protein profile it is meant the distribution of the molecular weights of the proteins that make up the protein crown.
The present method to assist in the early diagnosis of pancreatic adenocarcinoma which allows the identification of subjects in an early state of disease or at risk of pancreatic adenocarcinoma for whom it is necessary or appropriate to carry out second-level examinations by classifying a plasma sample of these subjects according to two or more levels of risk in which the classification of said sample in at least one of said levels of risk indicates the need or opportunity to carry out second-level examinations, comprising the following steps:
As indicated above, the set of reference samples is represented by samples coming from healthy individuals and samples of patients suffering from pancreatic adenocarcinoma. Said set of reference samples may be the set used and analyzed in the present description, or may be a set different from the one used in the present description, for which at least one discriminant function (ƒ(α1, α2, α3, . . . αn), g(α1, α2, α3, . . . αn)) and two or more bands of molecular weight (A1, A2 . . . An) are predetermined as defined hereinafter. The method of the invention can in fact comprise a step (passage) of predetermining at least one discriminant function and at least two bands of molecular weight preceding step e, on the protein profiles of the reference samples obtained, them also, as described in points a) to d) of the method claimed and in the detailed description of the invention. The method of the invention therefore allows to subdivide the analyzed patients on the basis of two or more levels of risk, in which the two basic levels are not at risk and at risk, with a sensitivity, specificity and overall correctness each greater than 75%, each preferably greater than 77%, or, even more preferably, each greater than 78%. Furthermore, levels intermediate between the two above-indicated ones can be identified, such as, e.g. low-risk and high-risk.
A sample's belonging to one of these levels of risk is determined as described in greater detail below.
In particular, the sensitivity and the overall correctness can also be each greater than 80%, or even each greater than 81%.
In the embodiment of the invention reported also in the examples, the observed final values of sensitivity, specificity and overall correctness are: 83.7%, 78.6%, 81.8%, respectively.
For the purposes of the present invention, e.g. for the above-reported values, the sensitivity, specificity and overall correctness of the method of the invention can be calculated on the basis of the number of correctly classified elements, as follows:
According to the present invention, and as indicated in the glossary, the first-level examination, in connection with the diagnosis of pancreatic adenocarcinoma comprises, is represented by the examination carried out by the method of the invention.
For the purposes of the carrying out of the method of the invention, said second-level examinations comprise one or more among abdomen CT, abdomen MRI, echoendoscopy, ERCP.
According to the invention, in point b) gold nanoparticles having an average diameter of 100 nm can be used. Also gold nanoparticles having an average diameter of 100 nm±20 nm are useful for carrying out the invention.
The incubation can be carried out, for instance, by mixing two identical or alike volumes of serum and of nanoparticles in suspension as provided by the manufacturer, and by incubating the whole (mixture) for a defined time and temperature.
Serum incubation with the gold nanoparticles can be carried out at a temperature between 35 and 40° C., preferably between 36 and 39° C., even more preferably at about 37° C.
The incubation time is preferably between 40 and 120 minutes, like e.g., a period of time from 50 to 70 minutes, or also a period of about one hour.
In a specific embodiment, said incubation can be carried out at about 37° C. for a period of time of about one hour.
In order to concentrate the nanoparticles, the serum can be subjected to centrifugation after the incubation period; the expert in the field will readily set the centrifugation times and rpms, which could be, in a merely indicative form, about 12-18 minutes at about 12000-15000 rpm. Preferably, the material previously incubated will be kept at temperatures of about 4° C.
Optionally, the proteins-nanoparticles complex, freed from the supernatant, could be subjected to one or more passages b′) of washing with a suitable solution to eliminate weakly bound proteins. A suitable solution is represented, merely by way of example, by saline phosphate buffer.
The above-described process leads to the obtainment, after the last centrifugation (the washing passages will consist in suspending the pellet in an appropriate solution, centrifugation, supernatant removal, resuspension, etc.) from gold nanoparticles coated with the protein crown.
According to the method of the invention, the obtained pellets are resuspended by centrifuging the product of point b) or of point b′) in a solution suitable to load the material suspended therein in a denaturing gel. In particular, the same solution normally is denaturing and contains, e.g., SDS. A passage of some minutes (8-13′) at boiling temperature of the water in which test tubes are placed, containing the solution in which the product of point b) or of point b′) is resuspended after centrifuging, leads to the detachment of the protein crown from the gold nanoparticles (point c) of the method), thereby allowing to isolate the protein crown and load it onto a polyacrylamide gradient gel in order to obtain a protein profile (point d) of the method).
The gel could, e.g., have a 4-20% polyacrylamide gradient.
The samples can be, e.g., suspended into 20 microliters of solution, and about 10 microliters/well can then be loaded onto the gel.
The gels containing the protein profile of the loaded samples are then subjected to image acquisition (e.g., by CCD camera, and appropriate image acquisition programs) and the image can then be processed, e.g. by bottom cleaning and subsequent processing by suitable codes in order to obtain a graphical representation, reported on abscissa and median, with intensity peaks, like that shown in
For instance, the gels can be transferred inside an image acquisition system. By way of example, a Chemidoc MP system (Bio-Rad) can be used, containing a CCD (Charge-Coupled Device) camera to capture images, preferably high-quality ones, in real time, and allowing to accurately position the sample.
Gel images are processed to obtain a protein profile for each sample of the measurement.
A representative image of gel is shown in
Preliminarily, a removal of the bottom signal is preferably carried out to prevent systematic variations of exposure or luminosity from affecting the resulting profiles.
According to one embodiment, for each image, the bottom subtraction technique via ball pivoting, row by row, is adopted. The ball radius is kept preferably constant for all processed images.
Formally, if I(x, y) represents the intensity sampled from the original image of a gel, the sampled image (x, y) is obtained after the bottom removal procedure:
I(x,y)→I′(x,y) (1)
In other words, each image so processed can be considered as a function I′ of two spatial variables (x and y), in which the shift y is related to the molecular weights of the proteins (see Eq. 3) and the intensity value is related to the amount of proteins.
A projection P of the two-variable function I′(x, y) on a plane orthogonal to the image represents the resulting protein profile for a generic sample (j):
P
j(y)=I′(x,y) (2)
Finally, the shifts y can be converted into molecular weights by comparing the position of the known proteins (first lane) with a nonlinear equation of the type
f(y)=a1*e(b1*y)+a2*e(b2*y) (3)
Where parameters a1, b1, a2, b2 are obtained via an experimental data fit procedure (position of bands corresponding to known proteins) by the functional relationship f. For any fit procedure the determination coefficient R2 can be calculated as follows:
R
2=1−Sr/St
Where Sr is the sum of the squares of the residues (each residue is defined as the difference between the experimental data and the value assumed by the fit function at the experimental data itself) and St is the sum of the squares of the differences between the experimental data and the mean of the experimental data.
Among the infinite possible functional relationships, the two-term exponential curve formalized in equation (3) represents a very reasonable, and above all effective choice, as it allows a very stable and versatile fit procedure. With this choice the determination parameter, R2, expressing the good quality of the procedure itself, settles on values not lower than 0.9999 (where 1 is the maximum reachable). Otherwise, R2 is about 0.9894 for single-term exponential curves, 0.9929 and 0.9499 for third- and second-degree polynomials, 0.9614 and 0.9915 for single- and double-term power laws.
It is understood that these values of the determination parameter R2 are related to the reference data set considered herein by way of example. If another data set were used, as described herein, e.g., adding new samples, or starting from a different set of samples, R2 values might vary.
Thus, at the end of the image processing procedure, for each sample a one-dimensional protein profile is obtained which expresses the distribution of the molecular weight of the proteins that make up the crown. There are commercial and free programs to process gel images, like, e.g., programs such as ImageJ (Mathworks) and Image Lab™ (Bio-Rad) or the like. Preferably, the programs could be designed or modified to allow a bottom subtraction, so as to guarantee reproducibility in comparing different gels.
The protein profiles obtained from gel electrophoresis experiments for a set of known reference samples, with the same modes described above with regard to the samples to be analyzed, are input in an appropriate database, so as to associate each sample (healthy or diseased) with its personal profile.
By way of example,
Sample classification and the entailed distinction between subjects not at risk and patients at risk or potentially diseased is, according to the invention, carried out through statistical analyses.
In particular, on the basis of the set of reference samples, the overall molecular weight interval is subdivided into two or more bands of molecular weight A1, A2 . . . An (n being an integer greater than 2) so that each peak—corresponding to a respective electrophoretic band—of each protein profile of the set of reference samples belongs to one and only one of said bands. Said bands are selected so as not to cut the peaks observable in the overall protein profile. For protein band determination, the profiles of the reference set are considered individually and the determination of the bands of molecular weight is univocal for the entire set of data under consideration.
In one embodiment of the invention said bands can be, for example, two or more among A1: 10-20 kDa; A2: 20-25 kDa; A3: 25-35 kDa; A4: 35-45 kDa.
The integral area values are used as input parameters for the subsequent computation, i.e. for a phase of classification of the subjects according to two or more levels of risk as defined herein. Said classification phase is, according to the invention, carried out through a linear discrimination analysis.
Said linear classification approach can be applied both in two dimensions (
The best pair is that maximizing the overall correctness of the test as described hereinafter.
Linear discrimination analysis is a statistical methodology well-known in the art, formalized by R. A. Fisher in 1936, and represents one of the first and major references for the development of additional procedures underlying modern machine learning and data mining. Hence, it is not deemed necessary to delve into the details of said methodological approach, it being certainly within the reach of an expert in the field.
According to the present invention at least one linear discriminant function ƒ(α1, α2), g(α1, α3, α4) was found maximizing the ratio of the differences between the specific single-class mean values, compared to standard deviations of the experimental variables. The result can be displayed graphically both as a straight line (in two dimensions,
According to a preferred embodiment of the invention, the discriminant function comprises a first function ƒ(α1, α2) represented by the equation of a straight line of the type:
ƒ(α1,α2)=r1+r2*α1+r3*α2=0
According to one embodiment, and on the basis of the set of reference samples used as described in the examples that follow, the coefficients of the line assume the following values:
ƒ(α1,α2)=−7.513+22.120α1+9.659α2=0
According to the example described herein, a, represents the (integral area value) value of integral area defined by the band of molecular weight A1 between 10 and 20 kDa, and α2 the integral area value in band A2 between 20 and 25 kDa.
According to the present invention, at least one predefined threshold value VS is set, with which to compare the discriminant value VDf of the discriminant function ƒ(α1, α2), calculated for the pair of integral area values α1 and α2 as defined above.
From comparison, the unknown plasma sample is classified as potentially pathological (also referred to as potentially diseased) or at risk if the discriminant value VDf is greater than zero, otherwise it is classified as patient not at risk. Of course, as already indicated, more than two levels of potentially pathological or at risk of developing the pathology can be envisaged, therefore carrying out a classification in more than two classes, simply by setting more than one threshold value.
Moreover, given its simplicity and non-invasiveness, the examination carried out with the method of the invention can be repeated at successive times on previously examined samples in order to confirm obtained results or identify patients previously not at risk, who have, over time, undergone modifications of the protein crown that may move them into the at risk or the potentially pathological class, on which it is therefore strongly advisable to carry out the second-level examinations as defined herein.
Finally, the sensitivity, specificity and overall correctness of the test were calculated on the basis of the number of correctly classified elements, as follows:
Analogously, the at least one discriminant function may also comprise a second function g(α1, α3, α4) represented by the equation of a plane of the type:
g(α1,α3,α4)=p1+p2*α1+p3*α3+p4*α4=0
According to one embodiment, and on the basis of the set of reference samples used as described in the examples that follow, the coefficients of the plane can assume the following values:
Function g(α1, α3, α4) will therefore assume the following form:
g(α1,α3,α4)=0.139+2.301α1−1.517α3−α4=0
According to the example described herein, a, represents the (integral area value) value of the integral area defined by the band of molecular weight A1 between 10 and 20 kDa, α3 the integral area value in band A3 between 25 and 35 kDa, and α4 the integral area value in band A4 between 35 and 45 kDa.
According to the present invention, at least one predefined threshold value VS is set, with which to compare the discriminant value VD9 of the discriminant function g(α1, α3, α4), calculated for the triplet of integral area values α1, α3 and α4 as defined above. From comparison, the unknown plasma sample is classified as pathological or at risk if the discriminant value VD9 is greater than zero, otherwise it is classified as not at risk. Of course, as already indicated, more than two levels of pathological or at risk can also be envisaged, therefore carrying out a classification in more than two classes, simply by setting more than one threshold value.
As indicated above, having established that a sample falls within the “at risk” classification (VDg and/or VDf greater than 0) additional intermediate levels of risk can be defined. A possible example, according to the present invention, is represented by the levels defined as ‘low-risk’ or ‘high-risk’. These additional levels can be identified in specific values of VDf and/or VDg; for instance: when the value of VDg and/or VDf is greater than 0 and 50.25 the sample can be further classified as “low-risk”, therefore with a risk lower than samples classified as “high-risk”. When the value of VDg and/or VDf is instead greater than 0.25 the sample can be classified as high-risk.
Not wishing to use the low- or high-risk definitions, the levels defined herein can also be defined as
Level 0 (not at risk)
Level 1 (at risk)
Level 2 (low risk, i.e., with a risk level lower than level 3)
Level 3 (high risk, i.e., with a risk level higher than level 2)
Finally, the sensitivity, specificity and overall correctness of the test were calculated on the basis of the number of correctly classified elements, as follows:
Through the proposed method, the following final values of sensitivity, specificity and overall correctness were obtained: 83.7%, 78.6%, 81.8%, respectively.
In the example described herein, a set of 77 known reference samples (28 samples from healthy volunteers and 49 from oncological patients suffering from pancreatic adenocarcinoma) was used, and data analysis on the indicated pair (α1, α2) returned the maximum value of overall correctness of the test (80%) among all possible pairings, therefore the abovementioned areas represent the best combination in a two-dimensional space.
Moreover, the classification carried out by considering a triplet of values returned the best values. By way of example, with reference to the examples reported below, it was observed that the overall correctness of the test improves of about 1.8% in passing from two to three dimensions.
Object of the invention is also a computer program comprising code adapted to carry out steps e) to h) as defined in claim 1 when executed on a computer.
Optionally, such a program can also comprise means for the processing of the discriminant function to be provided in point e, by inputting data obtained from a specific set of reference samples comprising healthy individuals and individuals suffering from adenocarcinoma.
Object of the invention is also a storage medium comprising the computer program as defined above.
Finally, object of the invention is a kit to assist in the early diagnosis of pancreatic adenocarcinoma that allows the identification of subjects in an early state of disease or at risk of pancreatic adenocarcinoma for whom it is necessary or appropriate to carry out second-level examinations comprising one or more aliquots of one or more reactants adapted to carry out steps a) and d) and optionally b′) and a storage medium according to the present description.
1. Plasma Production and Storage
Plasma taken and collected in test tubes containing K2EDTA protease inhibitors (test tubes TM BD P100 Blood Collection System (Franklin Lakes, N.J., 7 USA) is subjected to centrifugation twice for 10 minutes at 2500 rpm for human plasma isolation (
2. Protein Crown Formation, Isolation and Characterization
Subsequently, a volume of 50 microliters of gold nanoparticles (Sigma Aldrich, Milan, Italy) is incubated with 50 microliters of human plasma for 1 hour at 37° C. After incubation, samples are centrifuged at 14,000 rpm for 15 min at 4° C. and subsequently washed thrice with saline phosphate buffer to eliminate weakly bound proteins. This step of the procedure causes precipitation of a pellet made up of coated gold nanoparticles of the protein crown. Pellets thus obtained are resuspended into 20 microliters of 1× loading solution (made up of: 50 mM Tris-HCl, 0.01% bromophenol blue (BPB), 0.5% sodium-dodecyl-sulfate (SDS), 10% glycerol, 0.1 M dithiothreitol (DTT)) and subsequently boiled for 10 minutes at 100° C. This step allows to isolate the protein crown from the gold nanoparticles. Finally, for each sample of the measurement 10 microliters are loaded on a polyacrylamide gradient gel (4-20%) at the ends of which a difference of potential equal to 100 mV is applied for about 90 minutes. Gels are then washed in bidistilled water.
3. Image Acquisition and Processing
Gels are transferred inside an image acquisition system Chemidoc MP (Bio-Rad) containing a CCD (Charge-Coupled Device) camera to capture images in real time and allow to accurately position the sample, thus generating optimized image data. Gel images are processed through Matlab codes (Mathworks), written by the Inventors and expressly validated for electrophoresis data processing, in order to obtain a protein profile for each sample of measurement. A representative gel image is shown in
Formally, if I (x, y) represents the intensity sampled from the original image of a gel, the sampled image (x, y) is obtained after the bottom removal procedure:
I(x,y)−I′(x,y) (1)
In other words, each image can be considered as a function (I′) of two spatial variables (x and y), in which the shift y is related to the molecular weights of the proteins and the intensity value is related to the amount of proteins (see
The projection (P) of this two-variable function on a plane orthogonal to the image is the resulting protein profile for the generic sample (j) (
P
j(y)=I′(x,y) (2)
Finally, the shifts y can be converted into molecular weights by comparing the position of the known proteins (first lane) with a nonlinear equation (
f(y)=a1*e(b1*y)+a2*e(b2*y) (3)
Among the infinite possible functional relationships, the two-term exponential curve formalized in Eq. 3 represents a very reasonable, and above all effective choice, as it allows a very stable and versatile fit procedure. For that matter, the determination parameter R2, expressing the good quality of the procedure itself, settles on values not lower than 0.9999 (1 is the maximum reachable) using Eq. 3, but is about 0.9894 for single-term exponential curves, 0.9929 and 0.9499 for third- and second-degree polynomials, 0.9614 and 0.9915 for single- and double-term power laws. Anyhow, at the end of the image processing procedure, for each sample a one-dimensional protein profile (
4. Statistical Processing of One-Dimensional Protein Profiles and Blood Test Predictive Ability
The protein profiles obtained from gel electrophoresis experiments were input in an appropriate database, so that each subject (healthy or diseased) was associated with its personal adsorbed.
Sample classification and the entailed distinction between healthy subjects and oncological patients is carried out through statistical analyses. In particular, each profile is subdivided into bands of molecular weight, so that each electrophoretic band belongs to one and only one of said intervals.
The linear function maximizing the ratio of the differences between the specific single-class mean values, compared to standard deviations of the experimental variables, was found. The methodology is well-known and the result can be displayed graphically as a straight line (in two dimensions) or a plane (in three dimensions) subdividing the parameter space (the areas of the electrophoretic profiles) into two regions: that associated with the distribution of non-oncological samples and that corresponding to the counterpart of the subjects suffering from pancreatic tumor. The generic unknown sample is therefore classified based on its position in the parameter space of the areas. In the proposed case, the equation of the linear discrimination (straight) line (
ƒ(α1,α2)=−7.513+22.120α1+9.659α2=0 (4)
Where α1 represents the integral area defined between 10 and 20 kDa, and α2 the integral area in the 20-25 kDa range. A generic sample is therefore classified at risk, i.e., potentially diseased (potentially pathological) if the values of the areas of its electrophoretic profile meet the inequality f>0 in which the plasma sample is classified as potentially pathological (also referred to as potentially diseased) or at risk if the discriminant value VDf is greater than zero, otherwise it is classified as subject not at risk. Finally, sensitivity, specificity and overall correctness of the test were calculated on the basis of the number of correctly classified elements, as follows:
77 samples were used (28 samples from healthy volunteers and 49 from oncological patients suffering from pancreatic adenocarcinoma) and data analysis on the indicated pair (α1, α2) returned the maximum value of overall correctness of the test (80%) among all possible pairings, therefore the above-mentioned areas represent the best combination in a bidimensional space. Moreover, the extension to the three-dimensional case is analogous and returned better results. In particular, the equation of the plane as discriminant linear geometric entity is the following:
g(α1,α3,α4)=0.139+2.301α1−1.517α3−α4=0 (5)
Where α1 represents the integral area defined between 10 and 20 kDa, α3 the integral area in the 25-35 kDa range, and α4 the integral area in the 35-45 kDa range. The generic sample is therefore classified as at risk (potentially diseased) if g>0, otherwise it is considered not at risk. Through the proposed method, the following final values of sensitivity, specificity and overall correctness were obtained: 83.7%, 78.6%, 81.8%, respectively. Any unknown samples can then be classified through assessment of the electrophoretic profile, measurement of the specified integral areas and comparison with Eq. 5.
| Number | Date | Country | Kind |
|---|---|---|---|
| 102019000012555 | Jul 2019 | IT | national |
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/IB2020/056251 | 7/2/2020 | WO | 00 |
