Fig. 2: Network-guided rationalization of PPARA/PPARG as targets in IBD.

a An interactive web-based platform allows the querying of paths of gene clusters in the IBD map [ref. ⁹; see Supplementary Fig. 1] to pick high-value targets with a few mouse clicks and generate a comprehensive automated target ‘report card’. The components of a ‘target report card’ is shown (right): predicted ‘therapeutic index’ (likelihood of Phase III success), IBD outcome (prognostic potential in UC and/or CD), network-prioritized mouse model, estimation of gender bias, and predicted tissue cell type of action. b–h Components of a target report card for PPARA and PPARG are displayed. Bar plot (b) displays the rank ordering of normal vs. ulcerative colitis (UC) /Crohn’s Disease (CD) patient samples using the average gene expression patterns of the two genes: PPARG/PPARA. Samples are arranged from highest (left) to lowest (right) levels. ROC-AUC statistics were measured for determining the classification strength of normal vs IBD. Violin plots (b) display the differences in the average expression of the two genes in normal, UC, and CD samples in the test cohort that was used to build the IBD-map in⁹. Bar plots in panel c–d show the rank ordering of either normal vs. IBD samples (c) or responder vs. non-responder (R vs. NR; d), or active vs. inactive disease, or neoplastic progression in quiescent UC (qUC vs. nUC; d) across numerous cohorts based on gene expression patterns of PPARG and PPARA, from highest (left) to lowest (right) levels. Classification strength within each cohort is measured using ROC-AUC analyses. Bar plots in panel (e) show the rank ordering of either normal vs IBD samples across numerous published murine models of IBD based on gene expression patterns of PPARG and PPARA as in (d). ACT adoptive T cell transfer. Classification strength within each cohort is measured using ROC-AUC analyses. Bulk = whole distal colon; epithelium = sorted epithelial cells. Schematic in (f) summarizes the computational prediction of the cell type of action for potential PPARA/G-targeted therapy, as determined using Boolean implication analysis. GSEID# of multiple publicly available databases of the different cell types and colorectal datasets used to make sure predictions are cited. Red boxes/circles denote that PPARA/G-targeted therapeutics are predicted to work on monocytes/macrophages and crypt-top enterocytes. Computationally generated therapeutic index (see Methods) is represented as a line graph in (g). The annotated numbers represent Boolean implication statistics. PPARA and PPARG align with other targets of FDA-approved drugs on the right of threshold (0.1). Two FDA-approved targets (green; ITGB1, 0.046; JAK2, 0.032), two abandoned targets (red; SMAD7, 0.33; IL11, 0.16), PPARA (gray, 0.064), PPARG (gray, 0.04), and the threshold (black, 0.1) are shown in the scale. Box plot in panel h shows that the level of PPARA/G expression is similar in the colons of both genders in health and in IBD, and hence, PPARα/γ-targeted therapeutics are predicted to have little/no gender predilection. The diamond square is the mean, and the arrows around it are 95% confidence interval.