Divergence map of CRISPR fitness co-dependency and Perturb-seq transcriptional coupling across 26.2 million human gene pairs

This resource is a divergence map comparing two widely used public signals of gene–gene functional coupling. The first is CRISPR knockout-fitness co-dependency from DepMap (correlated Chronos gene-effect profiles across cell lines); the second is transcriptional-response coupling from genome-scale Perturb-seq (Replogle et al. 2022, correlated perturbation-response signatures). Across 7,244 shared genes (26,234,146 unordered gene pairs) the two signals are nearly independent (r = 0.05), yet both recover known protein complexes, and transcriptional coupling shows stronger enrichment for curated synthetic-lethal pairs (SynLethDB) than fitness co-dependency under dependence-aware inference. The map scores every gene pair on both signals and ranks them by divergence (transcriptional percentile minus fitness percentile), annotated with synthetic-lethality and shared-pathway (Reactome) status. It accompanies the manuscript "Transcriptional Coupling Shows Stronger Synthetic-Lethality Enrichment Than CRISPR Fitness Co-Dependency Across Two Public Perturbation Datasets" and includes the complete scored map (all 26.2 million pairs as Parquet, plus a top-100,000 CSV), processed annotations, the analysis code with a single reproduction command, random seeds, exact source-data versions, figures, and checksums. Data and documentation are released under CC BY 4.0; the analysis code under the MIT License.