A gene-level perturbation atlas of network consequences across cancer-relevant genes and therapeutic targets

A precomputed, queryable atlas of a genome-scale mechanistic signaling model's PREDICTED network-propagation consequences of gene-level gain-/loss-of-function perturbations for 182 cancer-relevant genes and therapeutic targets across nine cancer-hallmark behaviours (1,638 gene-behaviour combinations). Each combination stores a signed consequence and its top-ranked three-layer decomposition (per-input paired-flip attribution and top-3 loss-of-function-actionable downstream nodes, a subset with approved-agent annotations). Includes the atlas JSON, a JSON Schema, a data dictionary, derived CSV tables, portable reproduction code (run_all.sh; fixed seed), data-quality and clustering-stability validation, figures, and SHA-256 checksums. All values are model predictions in arbitrary, baseline-referenced units - not experimental measurements. Companion resource to the model method paper. Inputs are gene-level, not variant-level; single default polarity per gene.