ChemFlow data, source data, and predictions for drug-target interaction benchmarking and EGFR screening

This record contains the curated data, source-data tables, split files, prediction tables, metrics, and provenance files supporting the ChemFlow manuscript. The archive includes BioKG/TTD-derived inputs, node-masked fold definitions, post-2020 validation data, hard-negative evaluation data, BindingDB EGFR/BTK/JAK2 source and fold files, EGFR/MCE screening and docking-selection source data, processed MD replicate source data, ablation/KGE source tables, and final prediction/metric tables used for manuscript figures and supplementary analyses. Large trained model archives are included in this record. The public code repository is available at https://github.com/SwarnavaCB/ChemFlow. Third-party source datasets such as BioKG, TTD, BindingDB, and vendor library resources should also be cited through their original providers. Author-generated source tables, split manifests, predictions, metrics, and workflow provenance are provided to support reproducibility of the ChemFlow manuscript.