Ó³»­´«Ã½

Studying the genetic basis of human phenotypes involves two primary strategies. Model-system experiments generate interpretable gene networks but do not establish relevance to human disease. In contrast, statistical genetics identifies variant- and gene-level associations but cannot test mechanistic models. Here, we bridge these approaches by introducing NERINE, a hierarchical model-based rare variant association test that incorporates gene network topology while remaining robust to network inaccuracies. NERINE supports analysis of networks from established pathway databases and model-system screens. A comprehensive search across pathway databases reveals associations for breast cancer, cardiovascular diseases, and type 2 diabetes not detected by single-gene tests. Applied to experimental screen-derived networks in Parkinson's disease (PD), NERINE highlights autophagy-, vesicle-trafficking-, and protein-homeostasis-related gene modules. Genome-scale CRISPR interference (CRISPRi) screening in human neurons and NERINE converge on PRL, revealing an intraneuronal α-synuclein/prolactin stress response that may impact resilience to PD.