A predicted cancer dependency map for paralog pairs

Contains predictions for paralog pairs from "Systematic prioritisation of context-specific paralog pair vulnerabilities in cancer", Daldal et al 2026. Compared to previous versions, this version is smaller in terms of paralog pairs covered (33,419 pairs) but has much less missing data (>99% complete]. A total of 2,721 pairs were uniquely predictable in HeLa or PC9, because gene essentiality for these two cell lines derives from separate genome-wide CRISPR screens whose gene coverage differs from that of the DepMap library. These pairs were removed for this version, but remain available in earlier versions (the full model matrix in v1 and v2; the GIMap matrix in v2). All other scores remain the same. full_SLprediction_matrix.csv.zip Description: Contains prediction scores generated by the "Full" model.This is the primary analysis dataset used in the paper. full_SLprediction_matrix_GIMAP.csv.zip Description: Contains prediction scores generated by the "Full" model trained using scores from the GIMap scoring system. This is provided for reference for those interested in GIs scored with a multiplicative model. Data Structure * Rows (Index): Cell lines, identified by their DepMap ModelID. * Columns: Paralog gene pairs, formatted as `GeneA1_GeneA2` (e.g., `ARID1A_ARID1B`). * Values: Prediction scores ranging from 0 to 1. Higher values indicate a higher probability of the predicted SL interaction. Usage Notes * Prediction files are compressed (.zip) to reduce size. Please unzip to access the .csv matrix. * The first column in the CSV files serves as the index (DepMap ModelIDs). excluded_pairs.csv This contains the 2,721 pairs removed from the matrices, along with their predicted scores in HeLa and PC9.