Teams

Automated pipeline for generating residue-level lipid contact annotations from PDB structures. The workflow queries RCSB PDB for structures containing lipid chemical components (117 recognized codes), downloads original PDB files, calculates all-atom heavy-atom contacts at 4.0 Angstrom cutoff, clusters sequences at 30% identity using MMseqs2, and partitions into train/validation/test splits. Produces a complete dataset of 4,704 proteins with 8,055,325 annotated residues.