Fig. 2: GRIN retains biologically interrelated genes.
From: Analyses of GWAS signal using GRIN identify additional genes contributing to suicidal behavior
a Workflow for assessing the capacity of GRIN to identify biologically related genes from well-defined, “gold standard” gene sets (GO, OMIM, DisGeNET) mixed with random genes. Figure made with BioRender.com. b Receiver-operating characteristic (ROC) curve and precision-recall curves (PRC) for GRIN. The blue line represents the mean of GRIN performance over 100 sets of glutamatergic signaling genes (GO:0035249) intermixed with an equivalent number of random genes from the multiplex network. Gray values indicate maximum and minimum values over 100 sets, and dotted lines indicate classification values based on random chance. c Median area under ROC (AUROC) and PRC (AUPRC) values for 100 sets of random genes intermixed with a gold standard gene set for 30 distinct gold standard gene sets. d Median precision, recall, and specificity values for 100 sets of a gold standard gene set intermixed with an equivalent number of random genes, repeated for 30 distinct gold standard gene sets. e Violin and box plots of precision, recall, and specificity values for 100 sets of dopaminergic synaptic signaling (GO:0001963) gold standard genes intermixed with different ratios of random genes (Noise). Median precision: 0.970 (2:1), 0.767 (1:1), 0.568 (1:2), 0.360 (1:4), 0.175 (1:10). Median recall: 0.735 (2:1), 1.00 (1:1), 1.00 (1:2), 1.00 (1:4), 1.00 (1:10). Median specificity: 0.970 (2:1), 0.696 (1:1), 0.620 (1:2), 0.554 (1:4), 0.528 (1:10). Dotted lines indicate classification values based on random chance. f Precision, recall, and specificity values for 100 sets of four combined gold standard sets (dopaminergic synaptic signaling (GO:0001963), central nervous system myelination (GO:0022010), Alzheimer’s disease (OMIM #: 607822, 104300, 606889, 608907, 602192, 615590), and major depressive disorder (DisGeNET)) mixed with five ratios of gold standard to random genes. Median precision: 0.942 (2:1), 0.823 (1:1), 0.593 (1:2), 0.343 (1:4), 0.158 (1:10). Median recall: 0.746 (2:1), 0.881 (1:1), 0.955 (1:2), 0.993 (1:4), 1.00 (1:10). Median specificity: 0.909 (2:1), 0.813 (1:1), 0.672 (1:2), 0.524 (1:4), 0.470 (1:10). Dotted lines indicate classification values based on random chance.
