Datasets and pre-trained model weights for a multimodal molecular representation framework for drug repurposing — Application to KRAS G12C

This record contains the molecular datasets and pre-trained model weights used in "A Multimodal Molecular Representation Framework for Robust Similarity-Based Drug Repurposing: Application to KRAS G12C Inhibitor Identification" (submitted to Briefings in Bioinformatics). Contents: molecules_271.sdf (271 Adagrasib-similar compounds), molecules_validation.sdf (15,358-compound independent validation set, PubChem BioAssay), adagrasib.sdf (reference inhibitor, PubChem CID 138611145), training_graphs.pt (PyTorch Geometric graph dataset, ~5M molecules from ZINC22, used for GNN pre-training), and gnn_encoder.pt (trained GNN encoder weights). Companion code: https://github.com/samanthabove08/multimodal-drug-repurposing-kras