Darwin PBPK Platform - Training Datasets v1.0.0

Training datasets for Darwin PBPK Platform v1.0.0, including: - **consolidated_pbpk_v1.parquet**: Processed PBPK data for 44,779 compounds (ChEMBL + TDC + KEC) - **chemberta_embeddings_consolidated.npz**: ChemBERTa embeddings (768d, 44,779 molecules) - **molecular_graphs.pkl**: Molecular graphs in PyTorch Geometric format **Dataset Details:** - Total compounds: 44,779 - Train/Val/Test split: 80/10/10 (scaffold-based, zero leakage) - PBPK parameters: Fu, Vd, CL - Sources: ChEMBL, TDC (Therapeutics Data Commons), KEC **Related Software:** - Repository: https://github.com/agourakis82/darwin-pbpk-platform - Software DOI: https://doi.org/10.5281/zenodo.17536674