Cell type-specific coexpression networks funnel genetic risk into core genes

Genetic effects on common disease are mediated through many genes in multiple cell types, yet how networks connect genome-wide genetic risk to core disease genes remains unclear. One key limitation is that disease-related gene networks are likely cell type-specific, but these are poorly captured by bulk and pseudobulk coexpression networks. Here we used population-scale single-cell immune transcriptomes to decompose interindividual gene coexpression into cell type-shared and -specific networks, which have nearly opposite properties. Shared network hubs were enriched in constrained housekeeping genes, whereas cell type-specific hubs were dynamically expressed, rapidly evolving, and enriched in cell type-matched enhancers and transcription factors. Specific hubs enriched for immune disease GWAS heritability, Mendelian genes, and drug targets; these enrichments were absent or weak for non-immune disease and shared hubs. Gene-gene network edges enriched for causal trans regulation identified by trans eQTLs and perturb-seq, with shared hubs skewed to upstream regulators and specific hubs skewed to downstream disease-proximal genes. Specific hubs increased the relative success of genetically supported immune drug targets from 1.7-fold to 2.7-fold; for example, the lupus drug target BAFF is a monocyte hub connected to the Mendelian gene TREX1 and interferon-responsive genes. Overall, accounting for shared housekeeping networks reveals that cell type-specific networks funnel genetic risk into downstream and dynamic core disease genes.