Fig. 5: RSS-NET analyses of 18 complex traits and 38 regulatory networks. | Nature Communications

Fig. 5: RSS-NET analyses of 18 complex traits and 38 regulatory networks.

From: Modeling regulatory network topology improves genome-wide analyses of complex human traits

Fig. 5

a Clustering of 38 regulatory networks based on t-distributed stochastic neighbour embedding. Details are provided in Supplementary Fig. 11. b Similarity between a given tissue-specific PECA-based network and 394 CAGE-based networks for various cell types and tissues (a: adult samples; c: cell lines; f: fetal samples). The similarity between a PECA- and CAGE-based network is summarized by Jaccard indices of their node sets (x-axis) and edge sets (y-axis). To simplify visualization, only labels of top four CAGE-based networks with the highest edge similarity are shown for each PECA-based network. See Supplementary Fig. 12 for additional results. c Ternary diagram showing, for each trait, percentages of the “best” enrichment model (with the largest BF) as M11: θ > 0, σ2 = 0, M12: θ = 0, σ2 > 0 and M13: θ > 0, σ2 > 0 across networks. See Supplementary Table 4 for numerical values. Shown are 16 traits having multiple networks more enriched than the near-gene control. d Comparison of context-matched PECA-based (y-axis) and CAGE-based (x-axis) network enrichments on the same GWAS. Dashed lines have slope 1 and intercept 0. See Supplementary Fig. 14 for additional results. e Median proportion of genes with \({P}_{1}^{{\mathsf{bma}}}\) higher than reference estimates (\({P}_{1}^{{\mathsf{base}}}\) or \({P}_{1}^{{\mathsf{near}}}\)), among genes with reference estimates higher than a given cutoff. Medians are evaluated among 16 traits in c. See Supplementary Table 5 for numerical values. Overlap of RSS-NET prioritized genes (\({P}_{1}^{{\mathsf{bma}}}\ge 0.9\)) with genes implicated in f knockout mouse phenotypes47 and g human Mendelian diseases49,50. An edge indicates that a category of knockout mouse or Mendelian genes is significantly enriched for genes prioritized for a GWAS trait (FDR ≤ 0.1). Thicker edges correspond to stronger enrichments. To simplify visualization, only top-ranked categories are shown for each trait (f 3; g 2). See Supplementary Data 4, 5 for full results. Trait abbreviations are defined in Supplementary Table 1.

Back to article page