CAPLES: Model Artifacts, Datasets, and Benchmark Results

Model Artifacts for CAPLES (Co-Attentive Protein-Ligand Ensemble Scorer) This dataset contains the trained model weights, pre-computed data, and benchmark results required to run the CAPLES inference tool. Contents: P5_seed_*: The 5 trained ensemble model checkpoints (.pt files) and assay statistics. P2_outputs: Pre-trained scalers and PCA models for ligand featurization. esm2_t33_650M_UR50D: Local copy of the ESM-2 protein language model. ankh-base: Local copy of the AnkH protein language model. P0_outputs: Reconciled training data (LP-PDBBind). P4_optuna.db: Hyperparameter optimization database. Benchmark Analysis: Parity plots and results for CASF-2013 and CASF-2016. Usage: This archive is designed to be mounted directly into the CAPLES Docker container. Unzip this file and pass the folder path to the predict.py script using the --model_dir argument. Code Availability: The source code, Dockerfiles, and usage instructions are available on GitHub: https://github.com/MI-hassan-lab/CAPLES