HelixSeed: A MaSIF-Ready α-Helical Seed Library from the Human Proteome for Surface-Driven de novo Binder Design

The current release focuses on the human (Homo sapiens) reference proteome by integrating structural models from the AlphaFold Protein Structure Database (AFDB) and experimentally determined structures from the Protein Data Bank (PDB). In total, HelixSeed contains 19,715 human protein structure entries and 109,603 extracted α-helical segments curated using DSSP-based secondary-structure annotation and length/quality filtering (≥10 residues).

For each helix fragment, we provide standardized structure files and precomputed MaSIF-compatible molecular surface representations, including solvent-excluded surface meshes and mapped physicochemical features (e.g., hydropathy, hydrogen-bond potential, and electrostatics). Following the MaSIF preprocessing procedure, surfaces are further partitioned into geodesic patches (9 Å for MaSIF-site scoring and 12 Å for MaSIF-search fingerprinting), yielding 79,927,906 local surface patches with associated descriptors/fingerprints.