Clinical decision model

Abstract

An embodiment of the invention provides a method for determining a patient-specific probability of disease. The method collects clinical parameters from a plurality of patients to create a training database. A fully unsupervised Bayesian Belief Network model is created using data from the training database; and, the fully unsupervised Bayesian Belief Network is validated. Clinical parameters are collected from an individual patient; and, such clinical parameters are input into the fully unsupervised Bayesian Belief Network model via a graphical user interface. The patient-specific probability of the healing rate of an acute traumatic wound is output from the fully unsupervised Bayesian Belief Network model and sent to the graphical user interface for use by a clinician in pre-operative planning. The fully unsupervised Bayesian Belief Network model is updated using the clinical parameters from the individual patient and the patient-specific probability of the healing rate of an acute traumatic wound.

1 - 7 . (canceled) 
     
     
         8 . A method of generating a model predicting a personalized risk of disease for a subject, comprising:
 a) generating, by a processor, at least one fully unsupervised Bayesian Belief Network (BBN) model using data from a training database, wherein
 i) the training database comprises a set of reference clinical parameters obtained from a plurality of subjects having known disease outcomes, and 
 ii) the fully unsupervised BBN model comprises a directed acyclic graph including a plurality of nodes, wherein each node includes at least two bins, each bin representing a value range of a clinical parameter associated with that node, and wherein each of the nodes comprises data identifying at least one conditional dependence relationship between a known disease outcome and the clinical parameter associated with that node; 
   b) inputting a set of clinical parameters into the fully unsupervised BBN model, wherein the set comprises clinical parameters of a sample collected from an individual subject;   c) generating, by the processor, a predicted clinical outcome comprising a subject-specific probability of developing and/or recovering from a disease for the individual subject, using the fully unsupervised BBN model as a classifier, wherein the classifier is a machine learning system that compares the clinical parameters of the sample with the set of reference clinical parameters of the plurality of subjects having known disease outcomes; and   d) displaying and/or outputting the predicted clinical outcome.   
     
     
         9 . The method of  claim 8 , wherein the predicted clinical outcome displayed and/or output in step d) is provided in an interactive format using a graphical user interface configured to:
 allow a user to select a potential clinical outcome and/or clinical parameters for the sample input in step b), and to   update a probability distribution for each of the remaining variables used to generate the BBN model, in response to the user's selection of the potential clinical outcome and/or clinical parameters.   
     
     
         10 . The method of  claim 8 , wherein the fully unsupervised BBN model is generated without human-developed decision support rules. 
     
     
         11 . The method of  claim 8 , wherein the set of clinical parameters used in steps a), b) and/or c) comprises biomarker levels collected from at least one of serum or biopsy tissue, the biomarker levels including gene expression levels for an IP-10 gene, IL-6 gene, MCP-1 gene, IL-5 gene, and a RANTES gene. 
     
     
         12 . The method of  claim 8 , further comprising estimating an accuracy level of the subject-specific probability of developing and/or recovering from a disease, the accuracy level comprising at least one of model sensitivity, model specificity, positive and negative predictive values, and/or overall accuracy. 
     
     
         13 . The method of  claim 8 , wherein the disease comprises:
 a) a breast cancer;   b) a thyroid malignancy; or   c) a transplant glomerulopathy.   
     
     
         14 . The method of  claim 13 , wherein the disease comprises a breast cancer and the clinical parameters comprise one or more of a Gail model cutoff, mammogram results, MRI results, breast size, and personal history of breast disease. 
     
     
         15 . The method of  claim 13 , wherein the disease comprises a breast cancer and the clinical parameters comprise ultrasound data and results of a clinical breast examination. 
     
     
         16 . The method of  claim 13 , wherein the disease comprises a thyroid malignancy and the clinical parameters comprise functional status of a thyroid nodule, number of cervical lymph nodes, serum thyrotropin level, pre-operative diagnosis, nuclear medicine rating, age, and ethnicity. 
     
     
         17 . The method of  claim 13 , wherein the disease comprises a transplant glomerulopathy and the clinical parameters comprise gene expression levels for an ICAM-1 gene, IL-10 gene, CCL-3 gene, CD-86 gene, CCL-2 gene, CXCL-11 gene, CD-80 gene, GNLY gene, and PRF-1 gene. 
     
     
         18 . The method of  claim 13 , wherein the disease comprises a transplant glomerulopathy and the clinical parameters comprise gene expression levels for a CD40LG gene, IFNG gene, CD-28 gene, CXCL-10 gene, CCR-5 gene, CD-40 gene, CTLA-4 gene, TNF gene, CXCL-9 gene, CX3CR-1 gene, FOXP-3 gene, EDN-1 gene, CD-4 gene, TBX-21 gene, FASLG gene, C-3 gene, CD3E gene, CXCR-3 gene, and CCL-5 gene. 
     
     
         19 . The method of  claim 13 , wherein the disease comprises a transplant glomerulopathy and the clinical parameters comprise gene expression levels for a VCAM1 gene, MMP9 gene, Banff C4d gene, MMP7 gene, and LAMC2 gene. 
     
     
         20 . The method of  claim 13 , wherein the disease comprises a transplant glomerulopathy and the clinical parameters comprise gene expression levels for a TNC gene, S100A4 gene, NPHS1 gene, NPHS2 gene, AFAP gene, PDGF8 gene, SERPINH1 gene, TIMP4 gene, TIMP3 gene, VIM gene, SERPINE1 gene, TIMP1 gene, FN1 gene, ANGPT2 gene, TGFB1 gene, ACTA2 gene, TIMP2 gene, COL4A2 gene, MMP2 gene, COL1A1 gene, COL3A1 gene, GREM1-2 gene, SPARC gene, IGF1 gene, SMAD3 gene, HSPG2 gene, FN1 gene, ANGPT2 gene, TGFB1 gene, ACTA2 gene, THBS1 gene, CTNNB1 gene, FGF2 gene, TJP1 gene, FAT gene, CDH1 gene, SMAD7 gene, CD2AP gene, CDH3 gene, CTGF gene, ACTN4 gene, SPP1 gene, AGRN gene, VEGF gene, and BMP7 gene. 
     
     
         21 . A method for determining a patient-specific probability of impaired wound healing, said method including:
 a) generating, by a processor, at least one fully unsupervised Bayesian Belief Network (BBN) model using data from a training database, wherein
 i) the training database comprises a set of reference clinical parameters obtained from a plurality of subjects having known wound healing outcomes, the set of clinical parameters comprising gene expression levels for a plurality of genes, and 
 ii) the fully unsupervised BBN model comprises a directed acyclic graph including a plurality of nodes, each node comprising at least two bins with each bin representing a value range of a clinical parameter associated with that node, wherein each of the nodes comprises data identifying at least one conditional dependence relationship between the known wound healing outcomes and the clinical parameter associated with that node; 
   b) inputting a set of clinical parameters into the fully unsupervised BBN model, wherein the clinical parameters are associated with a sample comprising at least one of serum, wound effluent, or biopsy tissue collected from an individual subject;   c) generating, by the processor, a predicted clinical outcome comprising a subject-specific probability of impaired wound healing for the individual subject, using the fully unsupervised BBN model as a classifier, wherein the classifier is a machine learning system that compares the clinical parameters of the sample with the set of reference clinical parameters of the plurality of subjects having known wound healing outcomes; and   d) displaying and/or outputting the predicted subject-specific probability of impaired wound healing.   
     
     
         22 . The method of  claim 21 , wherein the predicted clinical outcome displayed and/or output in step d) is provided in an interactive format using a graphical user interface configured to:
 allow a user to select a potential clinical outcome and/or clinical parameters for the sample input in step b), and to   update a probability distribution for each of the remaining variables used to generate the BBN model, in response to the user's selection of the potential clinical outcome and/or clinical parameters.   
     
     
         23 . The method of  claim 21 , wherein the fully unsupervised BBN model is generated without human-developed decision support rules. 
     
     
         24 . The method of  claim 21 , further comprising estimating an accuracy level of the subject-specific probability of wound healing, the accuracy level comprising at least one of model sensitivity, model specificity, positive and negative predictive values, and/or overall accuracy. 
     
     
         25 . The method of  claim 21 , wherein the clinical parameters used in steps a), b) and/or c) comprises gene expression levels for an IP-10 gene, IL-6 gene, MCP-1 gene, IL-5 gene, and a RANTES gene. 
     
     
         26 . The method of  claim 21 , wherein the clinical parameters used in steps a), b) and/or c) comprises gene expression levels for an IL-1α gene, IL-10 gene, IL-2 gene, IL-3 gene, IL-4 gene, IL-7 gene, IL-8 gene, IL-10 gene, IL-12(p40) gene, IL-12(p70) gene, IL-13 gene, IL-15 gene, Eotaxin gene, IFN-γ gene, GM-CSF gene, MIP-60  gene, and TNFα gene. 
     
     
         27 . The method of  claim 21 , wherein the clinical parameters used in steps a), b) and/or c) further comprises levels of RNA transcripts and translation products of one or more genes selected from the group consisting: ACTA2, ACVR1, ADM, ALCAM, ANGPT 1, ANGPT 2, ANGPT 4, BAX, BCL2, BCL2L, 18S, 18S, CAV2, CCL1, CCL11, CCL17, CCL19, CCL 2, CCL 20, CCL22, CCL25, CCL27, CCL28, CCL3, COL3A1, COL4A1, COL4A3, CSSF1, CSF2 CSF3, CTGF, CX3CL1, CXCL1, CXCL10, CXCL11, CXCL12, FGF10, FGF11, FGF12, FGF13, FGF17, FGF2, FGF3, FGF5, FGF7, FGF8, FGF9, FIGF, IFNG, IGF1, IGF2, IGFBP1, IGFBP2, IGFBP3, IGFBP3, IGFBP4, IGFBP5, IGFBP6, IGFBP7, IL10, IL11, IL6, IL7, IL8, IL9, ITGA5, ITGAL, ITGAM, ITGB2, KDR, KITLG, LBP, LTA, MMP7, MMP8, MMP9, MPO, NCAM2, NFKB1, NFKB2, NOS2A, OSMR, PDGFA, PDGFB, PECAM1, SMAD6, SMAD7, SOCS1, SOCS3, SOCS5, STAT3, TEK, TGFA, TGFB1, TGFB2, TGFB3, TGFBR1, DCL2L2, BMP1, BMP15, BMP5, BMP3, BMP4, BMP5, BMP6, BMP7, BMP8A, BMP8B, CALCA, CALCB, CAV1, CCL4, CCL4L1, CCL4L2, CCL5, CCL7, CD14, CD4, CD40, CD40LG, CD83, CD8A, CD8B, COL18A1, COL1A1, CXCL13, CXCL2, CXCL5, CXCL9, ECGF1, EDN1, EGF, EGR1, EPO, FADD, FAS, FGF1, FLT1, FN1, GAPDH, GDF3, GDF5, MSTN, GDF9, HGF, HMGB1, IAPP, ICAM2, IFNB1, IL12A, IL13, ILLS, IL16, IL17A, IL18, IL1A, 1L1B, 1L2, IL3, IL4, IL5, MAPK14, MET, MMP1, MMP10, MMP11, MMP12, MMP13, MMP14, MMP15, MMP2, MMP24, MMP3, PF4, PLA2G4A, PTGS1, PTGS2, SELE, SELP, SERPINE1, SLPI, SMAD1, SMAD2, SMAD3, SMAD4, TIE1, TIMP1, TIMP2, TIMP3, TNC, TNF, TNFSF10, VCAM1, VEGFB, VEGFC, XCL1, and XCL2.