Patent Application
20230330121
Bronchiolitis obliterans syndrome (BOS) occurs in children after hematopoietic stem cell transplant (HSCT) and is associated with high morbidity and mortality. BOS is an irreversible, obstructive lung disease caused by a combination of inflammation and immune response resulting in permanent pulmonary function decline. BOS, a significant cause of morbidity and mortality in hematopoietic stem cell transplant (HSCT) recipients, is often associated with graft versus host disease (GVHD) and is likely underdiagnosed in earlier stages. The occurrence of BOS in HSCT recipients is ˜10% (with 22% 3-yr mortality rate in patients with GVHD). The extremely poor prognosis of BOS is a very negative clinical and emotional diagnosis for patients, given the expectation of a “cure” of their primary condition following transplantation. Currently, there are large gaps in knowledge regarding risk stratification, early disease identification and optimal treatment of BOS after HSCT. Due to the relative insensitivity of spirometry to detect early airway obstruction, especially in young children, HSCT patients are often referred for pulmonary evaluation only after significant respiratory decline, and as a result, BOS is often undiagnosed until the disease has progressed beyond the point where therapy will significantly improve the disease course and outcome. Disclosed herein are methods for personalized early diagnosis of BOS and evidence-based identification of candidate medications and treatments for the prevention and/or treatment of BOS.
Disclosed herein are compositions and methods for the treatment of bronchiolitis obliterans syndrome (BOS), in particular, in association with hematopoietic stem cell transplant (HSCT recipients. In certain aspect, the disclosed compositions and methods may be used for early identification of individuals likely to develop BOS such that optimal treatment may be provided.
This application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.
Those of skill in the art will understand that the drawings, described below, are for illustrative purposes only. The drawings are not intended to limit the scope of the present teachings in any way.
Unless otherwise noted, terms are to be understood according to conventional usage by those of ordinary skill in the relevant art. In case of conflict, the present document, including definitions, will control. Preferred methods and materials are described below, although methods and materials similar or equivalent to those described herein may be used in practice or testing of the present invention. All publications, patent applications, patents and other references mentioned herein are incorporated by reference in their entirety. The materials, methods, and examples disclosed herein are illustrative only and not intended to be limiting.
As used herein and in the appended claims, the singular forms “a,” “and,” and “the” include plural referents unless the context clearly dictates otherwise. Thus, for example, reference to “a method” includes a plurality of such methods and reference to “a dose” includes reference to one or more doses and equivalents thereof known to those skilled in the art, and so forth.
The term “about” or “approximately” means within an acceptable error range for the particular value as determined by one of ordinary skill in the art, which will depend in part on how the value is measured or determined, e.g., the limitations of the measurement system. For example, “about” may mean within 1 or more than 1 standard deviation, per the practice in the art. Alternatively, “about” may mean a range of up to 20%, or up to 10%, or up to 5%, or up to 1% of a given value. Alternatively, particularly with respect to biological systems or processes, the term may mean within an order of magnitude, preferably within 5-fold, and more preferably within 2-fold, of a value. Where particular values are described in the application and claims, unless otherwise stated the term “about” meaning within an acceptable error range for the particular value may be assumed.
There is a significant unmet medical need to identify patients at high risk for severe BOS after transplantation. Identification of patients at risk for severe BOS would allow clinicians to modify post-transplant care. Clinical functional surveillance would intensify for high-risk patients and immune modifying therapies could be initiated before full evolution of BOS or early intervention with lower risk therapies considered in the highest risk patients. Post-transplant BOS risk stratification would allow clinicians to select only those patients who would benefit from intensive prospective monitoring and timely clinical intervention. Risk stratification will also be economical, as use of expensive tests in patients with low risk for BOS can be avoided. Applicant's findings provide an immediate clinical benefit, because currently there are no diagnostic tools on the market to address individual susceptibility to HSCT associated BOS and there are no laboratory tests or clinical strategies for predicting risk to develop BOS after transplantation or trajectory of disease after BOS diagnosis.
In one aspect, the disclosed methods may be used for serially (monthly) post-transplant testing of patients by transplant physicians to determine risk for BOS and stratify screening and treatment procedures according to risk and predicted disease course. With nearly 20,000 transplants done each year in the United States, and more worldwide, this BOS prediction algorithm will have wide clinical utility. The healthcare costs associated with a diagnosis of BOS are also very high with increased hospitalizations, testing and aggressive interventions and complications resulting in a mean annual per patient cost that is $151,000 higher in adults in the first year after HSCT. The disclosed methods may be used for both adult and pediatric transplant patients, as BOS targeted diagnostic and risk prediction tools are currently not available.
Disclosed herein are methods employing biomarkers predictive of BOS prior to its onset, which may allow for a BOS specific prediction model. In turn, the disclosure provides a novel diagnostic tool for pre-emptive BOS risk stratification and prediction of disease trajectory.
In one aspect, a method for treating bronchiolitis obliterans syndrome (BOS) after transplantation in an individual who has undergone a hematopoietic stem cell transplant (HSCT) is disclosed. The method may comprise:
In one aspect, the one or more biomarker may be selected from one or more of Afamin, alpha-1-acid glycoprotein 1 precursor, alpha-1-acid glycoprotein 2 precursor, alpha-1-antichymotrypsin precursor, alpha-1B-glycoprotrein precursor, alpha-2-antiplasmin, alpha-2-macroglobulin, alpha-2-macroglobulin, angiotensinogen preprotein, antithrombin-III, antithrombin-III precursor, apolipoprotein A-I, apolipoprotein A-II, apolipoprotein A-IV, apolipoprotein L1, apolipoprotein M, beta-2-glycoproteing 1 precursor, carbonic anhydrase 1, CASP8-associated protein 2, clusterin preproprotein, Complement C1r subcomponent, complement C2, complement C3 preprotein, complement C4A (Rodgers blood group)-like preprotein, complement C4-B preprotein, complement C4-like preprotein, complement C5, complement component C6, complement component C6 precursor, complement component C7 precursor, complement component C8 gamma chain precursor, complement component C9 preproprotein, complement factor B preprotein, complement factor H-related protein 1, complement factor H-related protein 1 precursor, corticosteroid-binding globulin precursor, endothelial cell-specific molecule 1, precursor, gelsolin, hemopexin precursor, heparin cofactor 2 precursor, histone-lysine N-methyltransferase ASH1L, hypoxia up-regulated protein 1 precursor, hypoxia-inducible factor 1-alpha, hypoxia-inducible factor 3-alpha, integrin-linked protein kinase, inter-alpha-trypsin inhibitor heavy chain H2 precursor, inter-alpha-trypsin inhibitor heavy chain H3, inter-alpha-trypsin inhibitor heavy chain H3 preproprotein, inter-alpha-trypsin inhibitor heavy chain H4, inter-alpha-trypsin inhibitor heavy chain H4 precursor, interferon-induced protein with tetratricopeptide repeats 1B, kallistatin, kelch-like protein 2, kelch-like protein 22, kelch-like protein 5, kinesin-like protein KIF1A, kininogen-1 precursor, leucine-rich alpha-2-glycoprotein precursor, manganese-transporting ATPase 13A1, microtubule-associated tumor suppressor candidate 2, Mitogen-activated protein kinase 1, Mitogen-activated protein kinase 2, pigment epithelium-derived factor, plasma kallikrein preproprotein, plasma kallikrein, plasma protease C1 inhibitor precursor, plasminogen precursor, polymeric immunoglobulin receptor precursor, pregnancy zone protein, pregnancy zone protein precursor, protein AMBP preproprotein, protein KIAA0100 precursor, protein KIAA0100, prothrombin preproprotein, prothrombin, retinol-binding protein 4 precursor, retinol-binding protein 4, RRP12-like protein, RRP12-like protein, SAC3 domain-containing protein 1, serotransferrin precursor, serotransferrin, serum albumin preproprotein, serum amyloid A-1 protein preproprotein, serum paraoxonase/arylesterase 1 precursor, sorting nexin-8, steroid hormone receptor ERR2, vitamin D-binding protein precursor, vitamin D-binding protein, vitronectin precursor, protein RRP5 homolog, ubinuclein-1, ubinuclein-1, interleukin-13 receptor subunit alpha-1 precursor, DENN domain-containing protein 4C, protein FAM208B, DENN domain-containing protein 4C, centromere-associated protein E, ATP-binding cassette sub-family A member 13, intersectin-1, probable threonine-tRNA ligase 2, cytoplasmic, intersectin-1, complement C3 preproprotein, arf-GAP with GTPase, ANK repeat and PH domain-containing protein 2, and combinations thereof.
In one aspect, the one or more biomarker may be an isoform of the aforementioned biomarkers, and may be selected from afamin isoform X1, afamin isoform X2, afamin isoform X3, afamin precursor, alpha-1-acid glycoprotein 1 precursor, alpha-1-acid glycoprotein 2 precursor, alpha-1-antichymotrypsin precursor, alpha-1B-glycoprotein precursor, alpha-2-antiplasmin isoform a precursor, alpha-2-antiplasmin isoform X1, alpha-2-antiplasmin isoform X2, alpha-2-macroglobulin isoform a precursor, alpha-2-macroglobulin isoform b, alpha-2-macroglobulin isoform c, alpha-2-macroglobulin isoform X1, angiotensinogen preproprotein, antithrombin-III isoform X1, antithrombin-III precursor, apolipoprotein A-I isoform 1 preproprotein, apolipoprotein A-I isoform 2, apolipoprotein A-II preproprotein, apolipoprotein A-IV precursor, apolipoprotein L1 isoform a precursor, apolipoprotein L1 isoform b precursor, apolipoprotein L1 isoform c, apolipoprotein L1 isoform X2, apolipoprotein M isoform 2, apolipoprotein M isoform X1, beta-2-glycoprotein 1 precursor, carbonic anhydrase 1 isoform a, carbonic anhydrase 1 isoform b, carbonic anhydrase 1 isoform c, CASP8-associated protein 2, clusterin preproprotein, complement C1r subcomponent isoform 1 preproprotein, complement C1r subcomponent isoform 2 preproprotein, complement C1 s subcomponent isoform 1 preproprotein, complement C2 isoform 1 preproprotein, complement C2 isoform 2 precursor, complement C2 isoform 3, complement C2 isoform 4, complement C2 isoform 5, complement C3 preproprotein, complement C4A (Rodgers blood group)-like preproprotein, complement C4-A isoform 1 preproprotein, complement C4-A isoform 2 preproprotein, complement C4-B preproprotein, complement C4-B-like preproprotein, complement C5 isoform 1 preproprotein, complement C5 isoform 2, complement component C6 isoform X1, complement component C6 isoform X2, complement component C6 isoform X3, complement component C6 isoform X4, complement component C6 isoform X5, complement component C6 isoform X6, complement component C6 isoform X7, complement component C6 precursor, complement component C7 precursor, complement component C8 gamma chain precursor, complement component C9 preproprotein, complement factor B preproprotein, complement factor H-related protein 1 isoform X1, complement factor H-related protein 1 precursor, corticosteroid-binding globulin precursor, endothelial cell-specific molecule 1 isoform a precursor, gelsolin isoform h, gelsolin isoform X6, hemopexin precursor, heparin cofactor 2 precursor, histone-lysine N-methyltransferase ASH1L isoform X1, histone-lysine N-methyltransferase ASH1L isoform X2, hypoxia up-regulated protein 1 precursor, hypoxia-inducible factor 1-alpha isoform 1, hypoxia-inducible factor 3-alpha isoform X11, integrin-linked protein kinase isoform 1, inter-alpha-trypsin inhibitor heavy chain H2 precursor, inter-alpha-trypsin inhibitor heavy chain H3 isoform X1, inter-alpha-trypsin inhibitor heavy chain H3 isoform X2, inter-alpha-trypsin inhibitor heavy chain H3 preproprotein, inter-alpha-trypsin inhibitor heavy chain H4 isoform 1 precursor, inter-alpha-trypsin inhibitor heavy chain H4 isoform 2 precursor, interferon-induced protein with tetratricopeptide repeats 1B, kallistatin isoform 1, kallistatin isoform 2 precursor, kelch-like protein 2 isoform 1, kelch-like protein 2 isoform 2, kelch-like protein 22 isoform X4, kelch-like protein 5 isoform 2, kinesin-like protein KIF1A isoform 1, kinesin-like protein KIF1A isoform 2, kininogen-1 isoform 1 precursor, kininogen-1 isoform 2 precursor, kininogen-1 isoform 3 precursor, leucine-rich alpha-2-glycoprotein precursor, manganese-transporting ATPase 13A1, microtubule-associated tumor suppressor candidate 2 isoform b, Mitogen-activated protein kinase 1, Mitogen-activated protein kinase 2, pigment epithelium-derived factor isoform 2, plasma kallikrein isoform 1 preproprotein, plasma kallikrein isoform X1, plasma protease C1 inhibitor precursor, plasminogen isoform 1 precursor, polymeric immunoglobulin receptor precursor, pregnancy zone protein isoform X1, pregnancy zone protein isoform X2, pregnancy zone protein isoform X3, pregnancy zone protein isoform X4, pregnancy zone protein isoform X5, pregnancy zone protein isoform X6, pregnancy zone protein isoform X7, pregnancy zone protein isoform X8, pregnancy zone protein precursor, protein AMBP preproprotein, protein KIAA0100 isoform 1 precursor, protein KIAA0100 isoform 2 precursor, protein KIAA0100 isoform X1, protein KIAA0100 isoform X2, prothrombin isoform 1 preproprotein, prothrombin isoform 2, retinol-binding protein 4 isoform a precursor, retinol-binding protein 4 isoform b, RRP12-like protein isoform X1, RRP12-like protein isoform X2, RRP12-like protein isoform X2, SAC3 domain-containing protein 1 isoform a, serotransferrin isoform 1 precursor, serotransferrin isoform 2, serotransferrin isoform 3, serum albumin preproprotein, serum amyloid A-1 protein preproprotein, serum paraoxonase/arylesterase 1 precursor, sorting nexin-8, sorting nexin-8 isoform X1, steroid hormone receptor ERR2, steroid hormone receptor ERR2 isoform X1, steroid hormone receptor ERR2 isoform X2, steroid hormone receptor ERR2 isoform X3, steroid hormone receptor ERR2 isoform X4, steroid hormone receptor ERR2 isoform X5, vitamin D-binding protein isoform 1 precursor, vitamin D-binding protein isoform 3 precursor, vitamin D-binding protein isoform X1, vitronectin precursor, protein RRP5 homolog, protein RRP5 homolog isoform X1, ubinuclein-1 isoform X9, ubinuclein-1 isoform X8, ubinuclein-1 isoform a, ubinuclein-1 isoform b, interleukin-13 receptor subunit alpha-1 precursor, DENN domain-containing protein 4C isoform X2, protein FAM208B isoform 1, protein FAM208B isoform X9, DENN domain-containing protein 4C isoform 1, DENN domain-containing protein 4C isoform X1, protein FAM208B isoform X13, centromere-associated protein E isoform X1, ATP-binding cassette sub-family A member 13, ATP-binding cassette sub-family A member 13 isoform X1, intersectin-1 isoform X10, probable threonine-tRNA ligase 2, cytoplasmic isoform X1, intersectin-1 isoform 5, intersectin-1 isoform ITSN-1, intersectin-1 isoform X2, intersectin-1 isoform X3, intersectin-1 isoform ITSN-s, intersectin-1 isoform 4, intersectin-1 isoform 7, intersectin-1 isoform X11, complement C3 preproprotein, intersectin-1 isoform X4, intersectin-1 isoform X5, arf-GAP with GTPase, ANK repeat and PH domain-containing protein 2 isoform X2, DENN domain-containing protein 4C isoform X3 and combinations thereof.
The biological sample may be any biological sample in which the biomarkers may be detected. Exemplary biological samples may include one or more of blood, plasma, or serum. In one aspect, the individual is a pediatric patient (i.e., less than 18 years of age). In a further aspect, the individual has undergone an HSCT procedure. In yet further aspects, the individual is a pediatric patient who has undergone an HSCT procedure. It will be understood that the detecting of one or more biomarkers may be carried out at multiple timepoints. For example, a timepoint may be selected from 14 days post-HSCT, 30 days post-HSCT, 60 days post-HSCT, and 100 days post-HSCT in said individual.
In one aspect, the method may further comprise determining a lung function parameter in the individual. The lung function parameter determination may be used, in conjunction with the one or more biomarkers, to predict whether said individual is likely to develop BOS. The lung function parameter may include, for example, forced expiratory volume (FEV1) and ventilation defect percentage (VDP), and combinations thereof. In one aspect, the lung function parameter may be determined using hyperpolarized xenon-129 (129Xe) magnetic resonance imaging (MRI). In one aspect, a risk assessment determination may be determined, the risk assessment determination being based on one or more lung function measurement and the determination of a level of one or more biomarkers, which in turn may be used to determine the risk of the individual being likely to develop BOS.
It should be noted that, applying the methods disclosed herein, one of ordinary skill in the art will understand the detection and use of the biomarkers in determining a risk of BOS onset for an individual. That is, it will be understood that the biomarker profile for an individual biomarker or group of biomarkers can be used to predict BOS risk, and that one of ordinary skill will understand that a given biomarker or set of biomarkers will increase or decrease at different timepoints, and that such levels may be used as covariates with the disclosed algorithms to determine a risk level for an individual. Exemplary methods for using the biomarkers and algorithms are supplied herein.
In one aspect, the method may comprise administration of a treatment, wherein if said individual is identified as likely to develop BOS, said individual is administered one or more of a statin, a steroid with flovent, montelukast, azithromycin, ruxolitinib, sotalol, atorvastatin, pravastatin, and combinations thereof.
In a further aspect, a method for identifying an individual likely to develop GVHD or GVHD with BOS is disclosed. The method may comprise detecting one or more markers selected from plasma protease C1 inhibitor (C1INH2), integrin-linked protein kinase (ILK), Kelch-like protein 5 (KLHL5), kinesin-like protein 22 (KIF22), SAC3 domain-containing protein 1 (SAC3D1), RRP12-like protein (RRP12), manganese transporting protein Atpase 13a (ATP13A1), sorting nexin 8 (SNX8), and caspase 8 associated protein 2 (FLASH/CASP8A2). In one aspect, the one or more biomarker may be CASPA8P2, wherein if CASPA8P2 is increased in said individual, said individual may be identified as likely to develop GVHD and BOS, said individual being treated with one or more of a statin, a steroid with flovent, montelukast, azithromycin, ruxolitinib, sotalol, atorvastatin, pravastatin, and combinations thereof. In further aspects, the one or more biomarker may be one Factor 1, vitronectin C1 inhibitor, and combinations thereof. In further aspects, the one or more biomarker may be selected from an activator of HIF1a and an inhibitor of degradation of HIF1a. In a yet further aspect, the one or more biomarker may be increased IFN-γ signaling and expression of fibronectin binding integrins.
In certain aspects, an algorithm may be used to determine one or more of the objectives stated herein, in particular the likelihood of an individual to develop BOS (risk assessment), or to predict the severity thereof, such that appropriate treatment may be carried out. In this aspect, the algorithm may be as follows:
Y
ij=ρi(tij)+Ui+Wi(tij)+Zi
where Yij is the measure of lung function, (for example, FEV1), obtained for the ith patient at time point tij. The function μi (tij)=ƒ(tij)+ΣkXikθk represents mean lung function evolution for the ith patient, which includes spline formulation ƒ(tij) to characterize overall lung function progression as FD and encompasses covariates Xik with corresponding association parameters θik. The term U1, is assumed to follow a normal distribution with mean 0 and variance ω2, and provides patient-specific variability. Wi(tij) is meant to be a stochastic process reflecting imaging variation over time within an individual patient; integrated Brownian motion is used to depict this process, which has variance σ2. Finally, Zi represents normally distributed measurement error with mean 0 and variance τ2. Candidate molecular markers Bi1, . . . , Bim may be used as covariates and denote the association parameters as γ1, . . . γm to fit the model on smaller data. If marker and outcome data were available from all CFF-PR subjects, this new model could be called the “Full Model”. A so-called “Reduced Model” can be fit using detailed data from Applicant's prospective study cohort (n=130). This approach will be taken for modeling longitudinal R5 and other measures of lung function. In one aspect, the measure of lung function may be ventilation defect percentage (VDP) as described in Walkup et al, “Xenon-129 MRI detects ventilation deficits in paediatric stem cell transplant patients unable to perform spirometry,” Eur Respir J. 2019 may; 53(5): doi:10.1183/13993003.01779-2019.
In one aspect, the model may take into account observed lung function for the patient at each time point, such as the mean FEV1 evolution for each patient, and may encompass covariates with corresponding association parameters. The model may accommodate patient-specific heterogeneity between FEV1 (or other lung function) trajectories. Furthermore, a stochastic process may be used to reflect the “saw tooth” variation over time for individual patients; integrated Brownian motion may be used to depict this process. Normally distributed measurement error may be accounted for from the PFT. Data for the level of markers may then be added to the model. The covariance functions and estimation algorithm have been described13. The model may be implemented using the ‘1menssp’ package. Clinical/demographic covariates based on historical literature may be considered. To internally validate predictive accuracy, both 10-fold and leave-one-out cross-validation, may be performed in order to balance between bias and variance when making selections (R)14. This subset of the cohort may then have covariate data added to the model to check predictive accuracy; metrics here may include previously described MAPE, RMSE and MAD. Sensitivity, specificity and area under the receiver operating characteristic curve (AUC) may be assessed to provide a dynamic prediction model using patient-specific information on longitudinal plasma samples and pulmonary function testing.
In one aspect, the method may comprise detecting one or more, or two or more, or three or more, or four or more, or five or more, or six or more, or seven or more, or 8 or more, or 9 or more, or 10 or more, or 11 or more, or 12 or more, or 13 or more, or 14 or more, or 15 or more or 16 or more, or 17 or more, or 18 or more, or 19 or more, or 20 or more, or 21, or more, or 22 or more, or 23 or more, or 24 or more, or 25 or more, or 26 or more, or 27 or more, or 28 or more, or 29 or more, or 30 or more, or 31 or more, or 32 or more, or 33 or more, or 34 or more, or 35 or more, or 36 or more, or 37 or more, or 38 or more, or 39 or more, or 40 or more, or 41 or more, or 42 or more, or 43 or more, or 44 or more, or 45 or more, or 46 or more, or 47 or more, or 48 or more, or 49 or more, or 50 or more, or each protein (biomarkers) (or corresponding expression of each protein) in the protein (biomarker) set, wherein said method is predictive of risk of BOS in said individual. In certain aspects, the method may comprise detecting no more than 50, no more than 49, no more than 48, no more than 47, no more than 46, no more than 45, no more than 44, no more than 43, no more than 42, no more than 41, no more than 40, no more than 39, no more than 38, no more than 37, no more than 36, no more than 35, no more than 34, no more than 33, no more than 32, no more than 31, no more than 30, no more than 29, no more than 28, no more than 27, no more than 26, no more than 25, no more than 24, no more than 23, no more than 21, no more than 20, no more than 19, no more than 18, no more than 17, no more than 16, no more than 15, no more than 14, no more than 13, no more than 12, no more than 11, no more than 10, no more than 9, no more than 8, no more than 7, no more than 6, no more than 5, no more than 4, no more than 3, or no more than two proteins (biomarkers) in the protein (biomarker) set. It is intended that detection of a protein or protein expression level may be carried out using any method known in the art or hereafter developed which allows a determination or estimation of the relative expression or amount of a given protein.
In one aspect, the method may comprise the step of comparing said expression level to a control value to obtain a combined score and/or a risk probability score. A principle component or similar analysis that combines all the data from more than one or all of the markers may be used as well to generate a score. The combined score may be used to assess strength of association between the expression level of one or more of the aforementioned proteins, and the clinical parameter of interest. In one aspect, the clinical parameter may be lung function decline. The risk probability score may be used to predict the degree of risk that an individual will have or develop lung function changes or other clinical events that are of interest during the progression of cystic fibrosis.
In one aspect, the one or more clinical parameters may be selected from a clinical parameter, for example, FEV1, BMI, PE, number of hospitalizations, antibiotic status, infection status, and/or other clinical feature predictive of BOS risk. These parameters may be selected using statistical methods described herein.
In one aspect, the clinical parameter may be lung function decline, wherein an individual classified as being high risk for development of BOS is treated via more aggressive therapy and increased monitoring.
In one aspect, the sample may be blood, serum, urine, plasma, PBMCs, BALF, nasal and/or lower airway brushings, sputum, GI biopsies, lung explants, and combinations thereof. The sample may be obtained using routine methods known in the art. Multiple samples may be obtained over a period of time, for example, once every day, once every other day, once a week, once every two weeks, once every three weeks, once monthly, or once every two months, or once every three months, etc.
In one aspect, the detection step may be carried out using mass spectrometry. For example, electrospray/matrix-assisted laser desorption ionization mass spectrometry may be used, as described herein.
In one aspect, the methods described herein may be carried out via the use of a computerized device. For example, one or more of the combined score or a risk probability score for developing BOS may be calculated using a computer.
In one aspect, a probability score of developing BOS may be used to create a predictive model within a web browser, the computer having a graphical user interface (GUI) in which an end user can interactively explore the predictive model within a web browser. In one aspect, the end user may be, for example, a physician, patient, or patient guardian. The end user may use the interface to interactively explore the predictive model within a web browser. Predictions may be generated on an individual basis, utilizing data from cohorts, and the user may select which patient for which the prediction model is graphically illustrated. Inputs include, but are not limited to, clinical and demographic characteristics, such as those from electronic health records, and large-scale data from proteomics.
Measured and predicted lung function may be portrayed using one or more interactive graphs linked by a common timeline. One or more clinical parameters, biomarker status, and the risk of BOS may be displayed with corresponding confidence bands. The GUI may include patient-level and ecological descriptive variables as well as proteomic data that may be used to subset the pool of individual patients to select from, and the GUI may be expanded to include additional inputs. This may be used to facilitate, for example, the comparison of individually forecasted BOS development among individuals that are identical with respect to certain model covariates. Both static and temporal covariates may be displayed using graphical and text-based panels.
In certain aspects, the methods may be used in conjunction with evaluation of a drug or treatment for BOS. For example, a potential treatment may be administered to an individual in need thereof, and the disclosed methods may be carried out following administration of such drug or treatment.
The following non-limiting examples are provided to further illustrate embodiments of the invention disclosed herein. It should be appreciated by those of skill in the art that the techniques disclosed in the examples that follow represent approaches that have been found to function well in the practice of the invention, and thus may be considered to constitute examples of modes for its practice. However, those of skill in the art should, in light of the present disclosure, appreciate that many changes may be made in the specific embodiments that are disclosed and still obtain a like or similar result without departing from the spirit and scope of the invention.
Bronchiolitis obliterans syndrome (BOS) occurs in about 10% of children after hematopoietic stem cell transplant (HSCT) and is commonly, although not always associated with graft versus host disease (GvHD) (1). GvHD is frequent after HSCT, and most children with GvHD will not develop BOS. Currently the factors that determine whether BOS does or does not occur in GvHD are unknown. Mortality rates from BOS as high as 22% are reported and morbidity in children is very significant, in large part due to delayed diagnosis. In April 2018, the National Heart Lung and Blood Institute (NHLBI), the National Institute of Child Health and Human Development (NICHD) and the National Cancer Institute (NCI) convened a workshop of multi-disciplinary experts in pulmonary complications after HSCT in children. The executive summary of the meeting concluded that there are insufficient prospective and evidence-based studies to define the actual incidence, risk factors and biomarkers for BOS (2). Additionally, current approaches to understanding pediatric lung injury after HSCT are limited by small numbers of patients, the absence of agreed upon definitions, and management approaches to disease. In particular, the report concluded that studies of children are severely limited by the inability of many or most children to perform lung function testing with accurate results.
Currently, spirometry is the strategy of choice for detecting those at high risk of developing post BMT pulmonary complications. Several studies in adults have reported that changes in spirometry, including FEF25-75% and FEV1 in the first 3 to 6 months post-transplant can predict late onset non-infectious pulmonary complications (LONIPC) (3-6). The GvHD scoring system for detection of BOS incorporates lung function changes in FEV1, FEV1/FVC and FEF25-75% after HSCT (7). In pediatric practice, spirometry presents major challenges. Although spirometry may be accurate in children older than age 6 after considerable training, results are inconsistent. Children post HCST are often acutely ill and are unable to participate meaningfully in spirometry training. Consequently, although spirometry may be successful in other pediatric populations, such as children with cystic fibrosis, success rates are lower in children after HCST, even in teenagers (8). Lack of reliable spirometry in most children undergoing HSCT delays BOS diagnosis and lung damage incurred is often already irreversible when detected.
Plasma proteomics has been used to identify markers of response to therapy in early acute GvHD (9, 10) leading to development of valuable predictive algorithms (11-14). More limited analysis in BOS has identified MMP3 as a possible key marker (15). Although the depth of analysis was limited to −1,000 protein isoforms in plasma, the compelling results of these studies demonstrated the power of non-biased proteomics.
Given the limitations of existing diagnostic modalities in children, novel approaches are needed to facilitate early, accurate prognosis of BOS and timely treatment. Here, Applicant demonstrates that non-biased plasma proteomics may be used to identify aberrant pathway signaling and biomarkers of BOS more than 60 days before clinical diagnosis, and may be validated via ELISA. This approach may be used to identify key pathways in the progression of GvHD to BOS, facilitate earlier diagnosis, allow patient-personalized monitoring of response to therapy, and identify novel therapies or prophylactic medications.
Methods
Banked Sample Selection
The study design is schematically presented in
Mass Spectrometry
Applicant used a modified proteomic approach including methods previously developed (16-27) to analyze complex proteomes such as that of plasma. Applicant has developed mass spectrometry (MS)-based approaches for the identification and quantitation of thousands of proteins in 10 μl serum or plasma samples (see, e.g., U.S. patent application Ser. No. 15/927,575). Extremes of phenotype plasma samples from BMT pediatric patients banked soon after transplant and at the time of GvHD-BOS diagnosis were analyzed by MS as described before (16, 20-23, 28, 29). Briefly, whole protein from plasma was prepared using PureProteome Albumin/IgG magnetic beads and gel and column chromatography, followed by tryptic digestion. Following preparation of whole protein from the samples, protein peptides were extracted and subjected to data-dependent sequencing and MS analysis, as previously described. Samples loaded into an Ultimate 3000 HPLC autosampler system (Thermo Fisher, Inc.) were eluted by reverse-phase chromatography into an LTQ velos Pro mass spectrometer fitted with a nanospray ion source for analysis. The mass analyzer was set up for a data-dependent mode using dynamic exclusion settings: repeat count=2; repeat duration=10 seconds; exclusion list size=100; exclusion duration=30 seconds; exclusion mass width=1.5 amu. Collision-induced dissociation (CID) was used to fragment peptides, and CID spectra are searched against a human fasta database using the Proteome Discoverer™ software. For high stringency analyses, a reverse database decoy and percolator (30) were used to control for false discovery. A threshold filter of >1.5 for peptide XCorr score is used for sequence identification. For these studies, this typically produced a range of coverage for identified proteins of 5.16%-100% and an average of 50.86% for all proteins identified. This is a high stringency filter as 2 peptides of a protein can be sufficient for identification (21). High stringency data filtering allowed for more reliable quantitation using label free spectral counting. Protein quantitation was based on normalizing PSM for each protein identified by 2 or more peptide ions to total PSMs for all proteins identified by at least 2 peptide ions in each patient sample. For some pathway analyses the percolator algorithm (31) was used to evaluate proteomic datasets.
Pathway and Drug Effect Analysis
Applicant used MetaCore™ by Clarivate, version 20.1 build 70000 software (GeneGo Inc., Thomson Reuters) to map dysregulated proteins to GeneGO and GO pathways databases. Different algorithms were used to analyze the data including pathway maps, process network, GOProcesses, shortest path network analysis, and drug interaction analysis. MetaCore calculates p-values for networks generated by such algorithms by using hypergeometric distribution that assess the level of dataset matching to annotated pathways, networks, and drugs in the GeneGO databases.
ELISA Procedure and Analysis
Plasma samples from a larger cohort of >100 consecutive pediatric transplant recipients, including the cases tested by proteomics, collected 14, 30, 60 and 100 days after transplant were selected and ELISA performed for selected proteins shown to differ by diagnosis of GvHD and/or BOS. Undiluted plasma levels were measure by commercial ELISA kits according to manufacturer's instructions: ILK (Aviva, OKEH05402); C1INH (MyBioSource, MBS267814); CASP8AP2 (MyBioSource, MBS089197); Vitronectin (Thermo Fisher, EHVTN), ILK (Aviva, OKEH05402); CASP8AP2 (MyBioSource, MBS089197); SAC3D1 (MyBioSource, MBS9320606); IFIT1 (Aviva, OKEH02414); ATPase13 (MyBioSource, MBS282387) and KIAA0774 (MyBioSource, MBS9317669).
Results
Proteomic Discovery
Initial discovery MS analyses of a total of 14,107 BMT plasma protein isoforms found 12,629 (B vs. A), 12,726 (A vs. C), or 12,275 (B vs. C) isoforms were present in at least 50% of compared cohorts. Subsets of isoforms were differentially expressed (P<0.05): 1,263 in cohorts A vs. C (no GvHD vs. GvHD without BOS); 1,439 isoforms in cohorts A vs. B (no GvHD vs GvHD with BOS); and 359 isoforms in cohorts C vs. B (GvHD without BOS v GvHD with BOS). 509 isoforms are detected only in GvHD without BOS, and 499 only in GvHD with BOS.
Pathway Analysis: Prior to Diagnosis Timepoint
Pathway analysis at the prior-to-diagnosis timepoint revealed a significantly increased association with thrombophilia (p=1.56e-70, FDR=1.19e-67), complement activation (p=1.84e-35, FDR=1.15e-33), and angiotensin system maturation (p=2.47e-11, FDR=4.41e-10) in patients later diagnosed with BOS compared to controls without GvHD (cohort A vs. B). Associations were also seen with pulmonary fibrosis (p=5.14e-28, FDR=5.18e-27) and asthma (p=3.28e-24, FDR=2.56e-23) prior to diagnosis. Comparison between children without GvHD (cohorts A vs. B) and children with GvHD and BOS showed an association between BOS and reduced vitamin A metabolic pathways (p=1.28e-4, FDR=1.06e-2). Compared to patients with GvHD alone, patients with BOS exhibited altered cholesterol metabolism (p<e-7, differential expression of HDL proteins, APOA2 and APOA4) (cohort B vs. C).
Pathway Analysis: Time of Diagnosis Timepoint.
At time of diagnosis, compared to children without GvHD, the association of BOS with thrombophilia (p=1.53e-35, FDR=2.64e-33) and complement activation (p=7.21e-5, FDR=6.73e-4) remained, although with reduced statistical significance, while association with angiotensin system maturation remained strong (p=9.31e-14, FDR=7.54e-12). Similarly, iron transport and related inflammatory responses (p=8.31e-15, FDR=2.60e-12) were altered in BOS (cohort A vs. B). Children with GvHD and BOS also showed a continuing marked association with angiotensin system maturation (p=2.85e-19, FDR=7.13e-18) and dysregulated vitamin transport (p=3.68e-5, FDR=1.84e-2) compared to GvHD alone (cohort B vs. C).
Network Analysis: Prior to Diagnosis Timepoint
Markers clustered around several networks that predict features of GvHD without BOS and/or GvHD-BOS. In GvHD without BOS compared with no GVHD (cohort A vs C), for example, 2-fold or higher changes in plasma levels of integrin-linked protein kinase (ILK), Kelch-like protein 5 (KLHLS), kinesin-like protein 22 (Kid), SAC3 domain-containing protein 1 (SAC3D1), RRP12-like protein (PRP12), manganese transporting protein ATP-ase13A1 (ATP13A1), sorting nexin 8 (SNX8), and caspase 8 associated protein 2 (CASP8AP or FLASH) indicated alterations in a network (
MS proteomics detected differences in proteins that highly enriched process networks in both GvHD-BOS and GvHD vs. no GvHD (cohorts B and C vs. cohort A) including complement activation (p=7.75×10−39, FDR=3.79×10−37), Kallikrein-kinin signaling (p=4.47×10−17, FDR=1.09×10−15), blood coagulation (p=9.64×10−11, FDR=1.57×10−9), and platelet-endothelium-leucocyte interactions (1.12×10−6, FDR=1.37×10−5). The data suggest that proteins that segregate BOS from no GvHD cluster into coagulopathy, serine protease inhibition, vascular leak, angiotensin maturation and complement signaling, highlighting important clinical features of BOS such as inflammation and tissue injury (
Differential Complement Signaling
Nine complement factors are significantly elevated in prior to diagnosis samples (Table 3), identifying changes in complement pathway signaling (
Differential HIF1α Signaling
HIF1α activation was enriched in GvHD-BOS compared with no GvHD or GvHD alone. Samples banked from patients at the time of diagnosis timepoint showed elevation in the levels of the HIF1α activators MAPK1/ERK1 and MAPK2/ERK2 and HIF1α degradation inhibitor ubiquitin carboxyl-terminal hydrolase isozyme L1 (
ELISA Validation of Markers Discovered by MS
Cross-platform validation of the protein caspase 8 associated protein 2 (FLASH in
Drug Interactions with Altered Pathways
The proteomic data provided a detailed picture of the pathways implicated in the phenotypes of each examined cohort, indicating that statin and azithromycin therapies (Table 6) would be beneficial in reducing the lung disease exhibited by subjects that develop BOS. Furthermore, Applicant found an association of BOS with increased IFN-γ signaling and expression of fibronectin binding integrins in samples collected prior to diagnosis of BOS, identifying additional potential therapeutic targets.
Applicant found that MS proteomic identification of blood proteins may be used as potential biomarkers for the development of GvHD and/or GvHD-BOS. GvHD and BOS are closely inter-related, with most persons with BOS also having GvHD, although only a minority of persons with GvHD develop BOS. Applicant's analysis identified several processes and pathways that were abnormal in both GvHD and GvHD with BOS, including notable activation of complement in both at both timepoints studied. Moreover, a strong association with angiotensinogen maturation was also seen at both timepoints, in agreement with the frequent clinical observation of hypertension in children with GvHD and with BOS.
Currently it is not possible to distinguish which children with GvHD will or will not go on to develop BOS, and the mechanism of BOS initiation is poorly understood and studied. Applicant included a control group of children with GvHD but no BOS in addition to a control group of children with no GvHD to identify early biomarkers that distinguish children with GvHD destined to develop BOS from those who will not develop GvHD. Applicant's data show early alterations in vitamin A metabolic pathways and in cholesterol metabolism in patients with GvHD and BOS compared with children with GvHD but no BOS. Applicant's previous work showed an association between vitamin A levels and GvHD generally, supporting this finding (32).
Important progress has been made in identification of biomarkers predictive of response in early acute GvHD, and these have been incorporated into clinic practice. In this study Applicant used ELISA in a larger cohort of >100 consecutive HSCT, to validate changes in a subset of the markers identified in this study at multiple early timepoints after transplant (14, 30, and 100 days post-transplant), earlier than time of onset of BO. ELISA studies validated isoforms identified by proteomics as altered in GvHD, for example CASP8AP, supporting the discovery process and in some cases also or only in BOS. Applicant found that ATPase13A1 was reduced in children with GvHD who later developed BOS, but it was unchanged in those with GvHD and who did not develop BOS. IFIT1 was elevated in those with GvHD but no BOS, and unchanged in those with GvHD who later developed BOS, identifying candidate biomarkers specific for BOS that may be used for diagnosis and/or treatment of individuals.
Applicant used an in silico approach to mine proteomic data for potential pathways that could be therapeutic targets. Azithromycin, a medication commonly used to treat BOS and other causes of lung injury, was identified, supporting the validity of the Applicant's bioinformatic strategy. Additional therapies identified include statins and sotalol, which may be tested for both prophylaxis and treatment.
Early identification of BOS, prior to irreversible changes such as fibrosis will allow stratification of screening and treatment procedures according to risk and predicted disease course. Risk stratification permits modification of post-transplant care, and biomarkers may serve as effective surrogates for monitoring response. Clinical functional surveillance would intensify for high-risk patients and, immune modifying therapies could be initiated before full evolution of BOS, or very early intervention with lower risk therapies prior to clinical disease considered in the highest risk patients. The disclosed methods may also allow identification and testing of novel treatment and prevention strategies. Furthermore, biomarkers that forecast disease outcomes may be used for identification of novel therapeutic interventions to halt disease progression that can be tested in future multi-center trials.
11 Aziz M D, Shah J, Kapoor U, Dimopoulos C, Anand S, Augustine A, Ayuk F, Chaudhry M, Chen Y-B, Choe H K, Etra A, Gergoudis S, Hartwell M J, Hexner E O, Hogan W J, Kitko C L, Kowalyk S, Kroger N, Merli P, Morales G, Nakamura R, Ordemann R, Pulsipher M A, Qayed M, Reshef R, Rosier W, Schechter T, Schreiner E, Srinagesh H, WOlfl M, Wudhikarn K, Yanik G, Young R, Özbek U, Ferrara J L M, Levine J E. Disease risk and GVHD biomarkers can stratify patients for risk of relapse and nonrelapse mortality post hematopoietic cell transplant. Leukemia 2020; 34: 1898-1906.
16: 2342-2351.
Methods Mol Biol 2009; 544: 325-341.
Existing predictive pulmonary disease models successfully developed for cystic fibrosis and incorporate novel plasma protein biomarkers identified by mass spectroscopy (n=21) and validated by ELISA (n=130) along with existing pulmonary function data may be used to provide proof of concept for dynamic predictive modeling of BOS risk and disease progression in patients after HSCT.
Novel Predictive Modeling of Lung Disease.
Applicant has previously successfully used functional data (FD) analysis to characterize lung function at the CF population level and to accurately predict rapid decline in the individual patient8-12. Using this predictive algorithm13, the personalized behavior of biomarkers can be added to the lung function prediction algorithm and provide enhanced predictive probability of rapid decline. The model was applied to 88 subjects with FEV1 data and cross-sectional serum samples. Subjects were randomly split into training (80%) and test (20%) cohorts. Clinical/demographic covariates available from the preliminary sample collection included genotype (number of F508del alleles) and sex. Marginal testing using the Akaike information criterion (AIC), principal components regression and LASSO techniques (26) were each applied to examine a pilot panel of predictive proteomic markers. Marginal testing and LASSO indicated 114 markers demonstrated improved model fit with inclusion of markers, compared to a model without the markers (Likelihood ratio test, P<0.05 for all 114 markers).
Validation metrics assessed in the test cohort included mean absolute deviation (MAD), root mean-square error (RMSE), mean absolute percentage error (MAPE) and correlation between predicted values and observed values. Based on the FEV1 scale, validation metrics showed relatively small prediction error. MAPE, which measures forecast accuracy as percent difference between actual FEV1 and predicted FEV1, shows that there is relatively small error between projected and actual FEV1 values, and that MAPE was lower in models that included proteomic markers from the pilot panel, compared to MAPE in the model excluding proteomic markers. Correlation between observed and predicted values is excellent (above 0.80) and is significantly higher than presently available measures (range from 0.54-.0.71). The top 20 biomarkers, determined based on these validation metrics, improved FEV1 prediction accuracy by 3.4-11.3 percentiles. For example, the biomarker and lung function data from 87 of the 88 subjects predict the progression of lung function for the 88th subject. The addition of marker information improves algorithm fit to clinical measures of FEV1 (
Taken together, Applicant's innovative data indicate that existing models of pulmonary disease prediction in cystic fibrosis can be modified in combination with novel validated plasma biomarkers predictive of BOS to develop a dynamic predication model of BOS risk and disease progression.
Proteomic Discovery of Predictive Biomarkers of BOS
On the whole, proteomic analyses of samples collected from patients with GVHD-BOS, shortly after BMT, predicted increased lung disease, cellular adhesion, and fibrotic signaling predicted. These findings correlated with indication of lung function obtained by MR imaging (
The disclosed biomarkers allow a number of goals to be accomplished. First, the markers in trajectory algorithms that would help forecast the development of GvHD and GvHD-BO may be integrated. Second, the markers allow one to “see inside” the black box of lung injury in the youngest children who are unable to perform spirometry, and this is a pressing need in the field. Third, the analyses can identify potential therapeutic agents, such as the ones identified above, supporting the efficacy of this approach. The analysis also identified additional FDA approved drugs that could be tested in treatment or prevention of BO.
Applications of the approach have utilized data on 21 individuals aged 6.35-24.21 years who were “at risk” of developing bronchiolitis obliterans syndrome (BOS) and received care at a single bone marrow transplant center at Cincinnati Children's Hospital Medical Center from 2006 to 2020. Proteomic data for the application includes a total of 499 protein isoforms collected from individual serum samples. The longitudinal outcome was lung function measured as forced expiratory volume in 1 second of percentage predicted (FEV1% predicted).
Method
Data with a small sample size and large number of candidate explanatory variables are common in omics studies. The available variable selection tools yield a single model, but other sparse choices of explanatory features may fit virtually equally well. Cox and Battey (2017) introduced a semi descriptive analysis procedure called hyper-cube to obtain a small set of acceptable simple representations, which additionally allows the investigation of nonlinearities and interactions1. This method was implemented for different choices of traditional regression models, which are not appropriate for longitudinal data. In parallel, Applicant proposed novel target functions based on a non-stationary Gaussian linear mixed effects model to predict rapid cystic fibrosis lung function decline2. Empirical research shows that this model's covariance form (integrated Brownian motion3) and set of novel target functions outperform more traditionally applied approaches4. Applicant propose a modification of this approach by embedding it for hypercube estimation, in order to simultaneously i) predict real-time risk of developing BOS; ii) perform variable selection among candidate proteomic markers and demographic/clinical variables.
To explain the procedure, consider the analysis of data from n independent subjects on each of which a large number, p, of covariates is measured together with a single response variable, γ (the outcome). Unlike fitting a model that includes all p covariates and then shrinking variables to arrive at a subset of optimal predictors, the procedure begins with fitting a large number of regression analysis, and for each of the model Applicant only included a smaller number of variables, q. The hypercube approach from Cox and Battey may be embedded into the BOS prediction model as follows:
Intrinsic variables such as demographic/clinical characteristics may be included in all of the regression models described below according to content relevance/clinical adjudication.
One may arrange variables in a q×q or a q×q×q cube, or q×q×q×q hypercube, where q≤15. Extensions to five or more dimensions are possible and may be needed based on the size of p. Described are the cubic case (q×q×q). It is expected that some positions in the cube can be empty or if some rows, columns, and so on have more than q entries, so that there is no loss of generality in the restriction of p, say, to be a perfect cube.
The rows, the columns, and so forth of the cube form 3q2sets, each set includes q variables. One may fit the prediction model for each set of q variables.
For each such analysis, select a small number of variables for further analysis. The choice might be keeping two variables with the smallest p-values for the t-test of coefficients, two most significant, or any variables that has p-values below some given threshold.
Therefore, each of the explanatory variables has been tested three times (one may consider a cube in the explanation), in different set of explanatory variables. The variables that are never selected or only selected once will be excluded. If the number of variables selected twice or more is still high (˜100), these variables may be arranged to form a second phase hypercube, which has a lower dimension than the first phase. The same procedure will be repeated, until the number of remaining explanatory variables to roughly 10˜20, which are potentially explanatory variables that are selected.
Then for final set of selected variables, a more detailed analysis is performed. Their correlation matrix may be computed and for any pair of variables with a correlation greater than 0.9, the corresponding scatter plot may be examined Depending on the nature of the pair of variables, one may decide to omit one or to replace the pair by the average of their standardized values or to continue with both. For each of the selected variables a single squared term is added to the model, to investigate the possible significant squared-term. Similarly, the linear by linear interactions of the pairs of variables may also be investigated.
Then, the final step of the analysis is to identify very small sets of variables that give adequate fit. Suppose one finds all important variables along with quadratic and interaction terms, say one finds k variables in total, those are the candidate variables for further analysis. For a small k, a cautious approach is to fit all 2k sets models and reject those clearly inconsistent with the data. This might be done, by using likelihood ratio test against the model involving all k potential variables. The resulting set of models may comprise all comprehensive models.
The algorithm may be implemented by modifying code from existing packages and creating new code using R software5.
Hypercube Selection Results
Applying to the previously described biomedical data, Applicant first arranged the biomarker (variable) indices in a 5×5×5×5 hypercube to include all 499 biomarkers. There are some empty entries to form a perfect hypercube (since 5{circumflex over ( )}4>499). Applicant then fit the non-stationary Gaussian linear mixed effects model to the biomarkers, using the sets of variables indexed by each dimension of the hypercube; 280 variables were classified as significant at the alpha 0.05 level in at least 3 of the 4 analyses in which they appeared. The choice was keeping two variables with the smallest p-value, or any variable with a p-value below 0.05. Variables that were never selected or only selected once are removed. Applicant again arranged the corresponding variable indices in a 7×7×7 cube and repeated the procedure twice more, successively reducing the number of potential candidate explanatory variables to 66 and finally, the two proteins selected P603 and P1188 (codes of the proteins, see code table provided herein). The choices of statistical significance level and the dimension of the initial hypercube are not put forward as definitive; here significance tests are used informally as an aid to interpretation and are calibrated to decrease the number of candidate variables.
In the above analyses, issues have arisen with model fitting when the model includes at least one pair of highly correlated (Pearson's ρ>0.95) proteomic biomarkers. This results in discarding one of the proteomic biomarkers. To avoid this issue, one may exclude one of the highly correlated biomarkers from the initial set of variables before performing the analyses. To see how excluding one of the highly correlated pairs would affect the results, one may consider a sub-group analysis:
For instance, suppose, P1 and P2 are highly correlated biomarkers and the model fails whenever both are included as covariates.
Select the two highly correlated proteins (ensure their observed values are not exactly the same over all 21 patients).
Perform hypercube analysis when P1 is excluded from the model at the very beginning but P2 is included.
Do the same analysis when P2 is excluded and P1 is included in the initial model.
Check the results to see if there is any significant change in findings.
Here a sub-group of variables are selected, first 82 proteomic biomarkers among 499 to perform the above analysis, in which these variables are all not highly correlated. In first 82 variables, there are 8 highly correlated pairs, which are {P7 and P8}, {P15 and P16}, {P132 and P133}, {P182 and P183}, {P460 and P461}, {P523, P524, P525, and P526}, {P538, P539, P540, and P541}, and {P547 and P548}. Then two trials are considered:
Applicant performed the described hypercube procedure to both sets of biomarkers in Trials 1-2. The analyses in both trials resulted in the same final set of biomarkers, which include {P144, P173, P191, P195, P212, P374, P460, P508, and P544}. Hence, excluding one of the pair of highly correlated biomarker from the initial set of explanatory variables would not change the final selected set of biomarkers. The hypercube procedure was additionally performed to set of all 82 proteins and this resulted in selecting, {P144, P173, P195, P374, and P475}, although, this time the hyper-cube procedure has discarded some of the models where highly correlated biomarkers were included in the model together. The highlighted biomarkers overlapped with the findings from Trials 1-2.
Internal Validation Results from Marginal Testing
Applicant conducted cross-validation analyses to investigate predictive power of the 499 candidate biomarkers. There were 21 patients; 7 of whom developed BOS. These patients contributed total of 107 observations over time.
Applicant initially considered 80-20% cross-validation (where 80% of the subjects were used for model development and remaining 20% subjects were used for model validation). Receiver-operating characteristic (ROC) analyses were performed to assess predictive power and accuracy of the null and full model, where the null model is the prediction model that includes only the clinical and demographic characteristics of the patients while the full model additionally includes specific candidate proteomic biomarkers and their interactions with the time variable age. Both the null and full model yielded AUC and sensitivity values around 1. Applicant then considered a validation analysis where N=16 patients (4 of whom developed BOS) were used for model development, and the remaining N=5 patients (3 of whom developed BOS) were used as the internal validation cohort. This division of the cohort was chosen to study a group of patients with a similar amount of lung function decline yet mixture of BOS prevalence. This selection was meant to test robustness by increasing the difficulty of the approach to correctly predict BOS. Findings were similar to 80-20% cross-validation analyses. Finally, Applicant performed an overall cross-validation where all 21 subjects were used for both model development and validation. While the analysis indicated that none of the candidate proteomic biomarkers improved AUC or sensitivity, there were 161 proteins that improved specificity. Starting specificity median (range) 0.9855 (0.9667-1) for the null models, and the average per-proteomic biomarker improvement was 0.0145 (0.0145-0.0164).
Biomarker Code and Protein Name
Individual patient predictions from the proposed prediction algorithm for BOS: Predicting Real-Time Risk of BOS.
The case examples provided herein are for hypothetical individuals based on an application of Applicant's disclosed prediction model to longitudinal cohort data previously described herein. Described is an exemplary method for constructing the target function to predict real-time risk of BOS. To predict real-time risk of BOS for the ith patient at time tik, one may utilize the predicted lung function Yi(tik) of that patient at that time point;
Let the covariate history up to a given time t of each patient be represented as i(t)={Xi,(tij,γij):tij≤t}. Based on this history, one may build a predictive probability distribution for the proportion of change from baseline (Yi(bl), the first observed FEV1 for the ith subject) to Yi (tik) being above given a threshold at time tik;
where δ is the threshold for identifying BOS event, Yi(bl) is the baseline FEV1 for the ith subject, and pi(tik) is the predicted real-time risk of developing BOS.
The real-time risks, also referred to as the predictive probabilities, are calculated for the target function (equation (1) above) which is defined as a drop-in FEV1% predicted from the baseline excess of δ=10% or 20% during the follow up in this example. One may calculate these probabilities based on two such clinically relevant target functions: δ=10% and δ=20%. Shown are three examples of individual patient predicted trajectories from the BOS dynamic regression model, along with the 95% confidence interval (CI) for predictive values of longitudinal FEV1, corresponding to three hypothetical patients in
All percentages and ratios are calculated by weight unless otherwise indicated.
All percentages and ratios are calculated based on the total composition unless otherwise indicated.
It should be understood that every maximum numerical limitation given throughout this specification includes every lower numerical limitation, as if such lower numerical limitations were expressly written herein. Every minimum numerical limitation given throughout this specification will include every higher numerical limitation, as if such higher numerical limitations were expressly written herein. Every numerical range given throughout this specification will include every narrower numerical range that falls within such broader numerical range, as if such narrower numerical ranges were all expressly written herein.
The dimensions and values disclosed herein are not to be understood as being strictly limited to the exact numerical values recited. Instead, unless otherwise specified, each such dimension is intended to mean both the recited value and a functionally equivalent range surrounding that value. For example, a dimension disclosed as “20 mm” is intended to mean “about 20 mm.”
Every document cited herein, including any cross referenced or related patent or application, is hereby incorporated herein by reference in its entirety unless expressly excluded or otherwise limited. All accessioned information (e.g., as identified by PUBMED, PUBCHEM, NCBI, UNIPROT, or EBI accession numbers) and publications in their entireties are incorporated into this disclosure by reference in order to more fully describe the state of the art as known to those skilled therein as of the date of this disclosure. The citation of any document is not an admission that it is prior art with respect to any invention disclosed or claimed herein or that it alone, or in any combination with any other reference or references, teaches, suggests or discloses any such invention. Further, to the extent that any meaning or definition of a term in this document conflicts with any meaning or definition of the same term in a document incorporated by reference, the meaning or definition assigned to that term in this document shall govern.
While particular embodiments of the present invention have been illustrated and described, it would be obvious to those skilled in the art that various other changes and modifications may be made without departing from the spirit and scope of the invention. It is therefore intended to cover in the appended claims all such changes and modifications that are within the scope of this invention.
This application claims priority to and benefit of U.S. Provisional Application Ser. No. 63/061,334, entitled “Composition and Methods for the Treatment of Bronchiolitis Obliterans,” filed Aug. 5, 2020, the contents of which are incorporated herein in their entirety for all purposes.
This invention was made with government support under HL151588, HL141286, HL154105, and HL142210 awarded by the National Institutes of Health. The government has certain rights in the invention.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/US21/44587 | 8/5/2021 | WO |
| Number | Date | Country | |
|---|---|---|---|
| 63061344 | Aug 2020 | US |
