Vishal Patel, MD, PhD

Objective
To demonstrate the use of aggregated and de-identified electronic health record (EHR) data for multivariate post-marketing pharmacosurveillance in a case study of azathioprine (AZA).

To determine a molecular basis for prognostic differences in glioblastoma multiforme (GBM), we employed a combinatorial network analysis framework to exhaustively search for molecular patterns in protein-protein interaction (PPI) networks. We identified a dysregulated molecular signature distinguishing short-term (survival<225 days) from long-term (survival>635 days) survivors of GBM using whole genome expression data from The Cancer Genome Atlas (TCGA). A 50-gene subnetwork signature… Show more

To determine a molecular basis for prognostic differences in glioblastoma multiforme (GBM), we employed a combinatorial network analysis framework to exhaustively search for molecular patterns in protein-protein interaction (PPI) networks. We identified a dysregulated molecular signature distinguishing short-term (survival<225 days) from long-term (survival>635 days) survivors of GBM using whole genome expression data from The Cancer Genome Atlas (TCGA). A 50-gene subnetwork signature achieved 80% prediction accuracy when tested against an independent gene expression dataset. Functional annotations for the subnetwork signature included "protein kinase cascade,” “IκB kinase/NFκB cascade,” and “regulation of programmed cell death” – all of which were not significant in signatures of existing subtypes. Finally, we used label-free proteomics to examine how our subnetwork signature predicted protein level expression differences in an independent GBM cohort of 16 patients. We found that the genes discovered using network biology had a higher probability of dysregulated protein expression than either genes exhibiting individual differential expression or genes derived from known GBM subtypes. In particular, the long-term survivor subtype was characterized by increased protein expression of DNM1 and MAPK1 and decreased expression of HSPA9, PSMD3, and CANX. Overall, we demonstrate that the combinatorial analysis of gene expression data constrained by PPIs outlines an approach for the discovery of robust and translatable molecular signatures in GBM. Show less

In this chapter, we begin by introducing a specific type of cancer – colorectal cancer – and the bioinformatic tools applied in the analysis of colorectal tumor specimens. We then introduce an emerging approach for analyzing multidimensional datasets, and we refer to the resulting construct as a “molecular subsystem.” We describe relevant issues related to the construction, interpretation, and validation of these subsystems. In particular, we provide a worked example at the end of chapter to… Show more

In this chapter, we begin by introducing a specific type of cancer – colorectal cancer – and the bioinformatic tools applied in the analysis of colorectal tumor specimens. We then introduce an emerging approach for analyzing multidimensional datasets, and we refer to the resulting construct as a “molecular subsystem.” We describe relevant issues related to the construction, interpretation, and validation of these subsystems. In particular, we provide a worked example at the end of chapter to illustrate the statistical concerns in constructing a molecular subsystem from proteomic data. Show less

We sought to examine the frequencies and patterns of nephrotoxicity and neutrophilia due to azathioprine (AZA), and to develop a prototype method for using large de-identified electronic health record (EHR) data to aid in post-market drug surveillance. We leveraged a de-identified database of over 10 million patient EHRs to construct a network of comorbidities induced by administration of AZA, where comorbidities were defined by baseline-controlled laboratory values. To gauge the significance… Show more

We sought to examine the frequencies and patterns of nephrotoxicity and neutrophilia due to azathioprine (AZA), and to develop a prototype method for using large de-identified electronic health record (EHR) data to aid in post-market drug surveillance. We leveraged a de-identified database of over 10 million patient EHRs to construct a network of comorbidities induced by administration of AZA, where comorbidities were defined by baseline-controlled laboratory values. To gauge the significance of the identified disease patterns, we calculated the relative risk of developing a comorbidity pair relative to a control cohort of patients taking one of 12 other anti-rheumatic agents. Nephrotoxicity as gauged by elevations in creatinine was present in 11% of patients taking AZA, and this frequency was significantly higher than in patients taking other anti-rheumatic agents (RR: 1.2, 95% CI: 1.04-1.43). Neutrophilia was highly prevalent (45%) in the population and was also unique to AZA (RR: 1.2, 95% CI: 1.17-1.28). Using a comorbidity network analysis, we hypothesized that the joint consideration of anemia (hemoglobin 190 IU/L) may serve as a predictor of impending renal dysfunction. Indeed, these two laboratory values provide approximately 100% sensitivity in predicting subsequent elevations in creatinine. Furthermore, the predictive power is unique to AZA, for jointly considering anemia and an elevated LDH provides only 50% sensitivity in predicting creatinine elevations with other anti-rheumatic agents. Our work demonstrates that the construction of comorbidity networks from de-identified EHR data sets can provide both sufficient insight and statistical power to uncover novel patterns and predictors of disease. Show less

Bimodal patterns of expression have recently been shown to be useful not only in prioritizing genes that distinguish phenotypes, but also in prioritizing network models that correlate with proteomic evidence. In particular, subgroups of strongly coexpressed gene pairs result in an increased variance of the correlation distribution. This variance, a measure of association between sets of genes (or proteins), can be summarized as the bimodality of coexpression (BiC). We developed an online tool… Show more

Bimodal patterns of expression have recently been shown to be useful not only in prioritizing genes that distinguish phenotypes, but also in prioritizing network models that correlate with proteomic evidence. In particular, subgroups of strongly coexpressed gene pairs result in an increased variance of the correlation distribution. This variance, a measure of association between sets of genes (or proteins), can be summarized as the bimodality of coexpression (BiC). We developed an online tool to calculate the BiC for user-defined gene lists and associated mRNA expression data. BiC is a comprehensive application that provides researchers with the ability to analyze both publicly available and user-collected array data. Show less

Colon cancer is driven by mutations in a number of genes, the most notorious of which is Apc. Though much of Apc's signaling has been mechanistically identified over the years, it is not always clear which functions or interactions are operative in a particular tumor. This is confounded by the presence of mutations in a number of other putative cancer driver (CAN) genes, which often synergize with mutations in Apc. Computational methods are, thus, required to predict which pathways are likely… Show more

Colon cancer is driven by mutations in a number of genes, the most notorious of which is Apc. Though much of Apc's signaling has been mechanistically identified over the years, it is not always clear which functions or interactions are operative in a particular tumor. This is confounded by the presence of mutations in a number of other putative cancer driver (CAN) genes, which often synergize with mutations in Apc. Computational methods are, thus, required to predict which pathways are likely to be operative when a particular mutation in Apc is observed. We developed a pipeline, PETALS, to predict and test likely signaling pathways connecting Apc to other CAN-genes, where the interaction network originating at Apc is defined as a "blossom," with each Apc-CAN-gene subnetwork referred to as a "petal." Known and predicted protein interactions are used to identify an Apc blossom with 24 petals. Then, using a novel measure of bimodality, the coexpression of each petal is evaluated against proteomic (2 D differential In Gel Electrophoresis, 2D-DIGE) measurements from the Apc1638N+/-mouse to test the network-based hypotheses. The predicted pathways linking Apc and Hapln1 exhibited the highest amount of bimodal coexpression with the proteomic targets, prioritizing the Apc-Hapln1 petal over other CAN-gene pairs and suggesting that this petal may be involved in regulating the observed proteome-level effects. These results not only demonstrate how functional 'omics data can be employed to test in silico predictions of CAN-gene pathways, but also reveal an approach to integrate models of upstream genetic interference with measured, downstream effects. Show less

Colorectal cancer progresses through an accumulation of somatic mutations, some of which reside in so-called “driver” genes that provide a growth advantage to the tumor. To identify points of intersection between driver gene pathways, we implemented a network analysis framework using protein interactions to predict likely connections – both precedented and novel – between key driver genes in cancer. We applied the framework to find significant connections between two genes, Apc and Cdkn1a… Show more

Colorectal cancer progresses through an accumulation of somatic mutations, some of which reside in so-called “driver” genes that provide a growth advantage to the tumor. To identify points of intersection between driver gene pathways, we implemented a network analysis framework using protein interactions to predict likely connections – both precedented and novel – between key driver genes in cancer. We applied the framework to find significant connections between two genes, Apc and Cdkn1a (p21), known to be synergistic in tumorigenesis in mouse models. We then assessed the functional coherence of the resulting Apc-Cdkn1a network by engineering in vivo single node perturbations of the network: mouse models mutated individually at Apc (Apc1638N+/−) or Cdkn1a (Cdkn1a−/−), followed by measurements of protein and gene expression changes in intestinal epithelial tissue. We hypothesized that if the predicted network is biologically coherent (functional), then the predicted nodes should associate more specifically with dysregulated genes and proteins than stochastically selected genes and proteins. The predicted Apc-Cdkn1a network was significantly perturbed at the mRNA-level by both single gene knockouts, and the predictions were also strongly supported based on physical proximity and mRNA coexpression of proteomic targets. These results support the functional coherence of the proposed Apc-Cdkn1a network and also demonstrate how network-based predictions can be statistically tested using high-throughput biological data. Show less