Skip to main content
. 2019 Oct 10;47(22):e146. doi: 10.1093/nar/gkz868

Figure 3.

Figure 3.

Pathophysiological relevance of prediction outcomes. (A) Results of feature importance analysis. The MDG score (x-axis) and its P value (y-axis) for individual features with significant (P < 0.05) disease related features highlighted in red (left panel). Enrichment of the significant features in different categories (neural, immune, circulatory, or digestive) as estimated by the hypergeometric test (right panel). (B) Network of neural features that were important in each disease model. The small nodes represent neural features that showed a significant (P < 0.05) MDG in the prediction model of the connected disease. The significant neural features, including those related to fetal brain (blue), astrocytes (red), and dorsolateral prefrontal cortex (green), were mapped to the relevant disorder (yellow). (C) Enrichment of positive calls for autoimmune diseases for the BLUEPRINT epigenome data for various immune cell types. Positive calls are more enriched in the regulatory regions of lymphocyte lineages rather than granulocytes. Shown here are regulatory regions marked by H3K4me1. The H3K27ac data is provided in Supplementary Figure S5.