We employed gene-network analysis to compare the biological processes underlying Externalizing, Behavioral Disinhibition, and SUDs. Using a random-walk34 algorithm within the PCNet2.035 interactome, we generated gene-networks for each latent genomic factor. As described previously17,52, network propagation was computed using NetColoc34, using an alpha=0.5, and a z-score cutoff of 3. We generated a multiscale systems map of the for each network using the Hierarchical community Decoding Framework (HiDeF v1.1 Beta) algorithm as implemented in Cytoscape’s Python package CDAPS53, using a maximum resolution of 2, and enrichment for GO terms limited to terms with 50-1,000 annotations, a minimum of 3 genes overlapping, and enrichment two sided p-value<0.05. All significant annotations for each community were manually aggregated into a single label for each community. GWAS Catalog enrichment was calculated using Experimental Factor Ontology based phenotype grouping for the GWAS catalog, downloaded from https://www.ebi.ac.uk/gwas/docs/file-downloads on 7 August 2024. Enrichment was calculated via hypergeometric test against all PCNet2.0 nodes, and Bonferroni corrected for all traits in the Experimental Factor Ontology.
