A Reproducible Comparison of Cancer Gene Dependency Rankings Across Archived DepMap Releases (23Q4 vs. 24Q4)
收藏资源简介:
An author-documented, reproducible computational stability benchmark quantifying the consistency of CRISPR-Cas9 cancer gene dependency scores between two major archived public releases of the Cancer Dependency Map (DepMap): DepMap 23Q4 Public (DOI: 10.25452/figshare.plus.24667905.v2) and DepMap 24Q4 Public (DOI: 10.25452/figshare.plus.27993248.v1). Creator & Institutional Affiliation: Rashid Kadayil, Pondicherry University (ORCID: 0009-0009-6398-4557). Resource Classification & Verification Scope: This archive is a derived computational release-comparison resource evaluating score stability across Broad Institute public datasets. It does not contain original wet-lab experimental screening data, does not constitute clinical or pharmacological target validation, and is not a peer-reviewed journal paper. The top-level MANIFEST_SHA256.txt hashes all 37 content files comprising this 38-entry release archive. Key Empirical Findings (Across 1,096 Matched Cancer Models and 17,916 Common Genes / 19.46M Evaluable Measurement Pairs): Global Score Concordance: Median Absolute Difference (MAD) = 0.0353 score units; Mean Absolute Difference = 0.0466; 95th percentile difference = 0.1289. Sampled pairwise Pearson r = 0.9884, Spearman rho = 0.9426 (N = 100,000 pairs, seed 42). Mean per-model Spearman rho = 0.9405 across all 1,096 cell lines. Candidate Target Prioritization Stability (Top-K Overlap): Top-20 candidate mean Jaccard index = 80.76% (mean 2.20 displaced genes); Top-50 mean Jaccard = 82.65% (mean 4.83 displaced genes); Top-100 mean Jaccard = 84.85% (mean 8.29 displaced genes). Boundary Dynamics (-0.50 cutoff): 170,190 paired observations (0.874% of valid universe) cross the nominal -0.50 dependency threshold, with 89.23% (151,859 crossings) originating in the borderline zone [-0.60, -0.40]. Gene Harmonization: Matches genes strictly by stable NCBI Entrez IDs, retaining 71 genes whose official symbols changed between releases. Package Contents & Reproducibility Infrastructure: Contains 5 harmonized CSV tables (results/), 3 publication-grade figures (figures/), complete modular Python execution pipeline (scripts/), automated test suite (tests/, 10/10 PASS), pre-executed Jupyter notebook with inline figures (depmap_pipeline.ipynb), and full remote Linux container execution transcript (EXECUTION_TRANSCRIPT.log; verified on Kaggle CPU, Version 16 in 103.1s).



