Molecular descriptor profiling and multi-tool prediction of peptide hemolytic potential

This dataset comprises 469 unique peptide sequences ranging from 7 to 35 amino acids, compiled for computational characterization of putatively non-hemolytic peptides. Each peptide record includes sequence-derived physicochemical, conformational, membrane-interaction, aggregation, and functional-prediction parameters. The annotated descriptors include peptide length, net charge, isoelectric point, normalized hydrophobicity, normalized hydrophobic moment, amphiphilicity index, disordered-conformation propensity, polyproline-II coil propensity, linear moment, membrane-penetration depth, tilt angle, angle subtended by hydrophobic residues, and propensity for in vitro aggregation. Potential biological functions were evaluated using established peptide-prediction tools. Antimicrobial potential was assessed with AMPScannerV2, cell-penetration probability with MLCPP, and hemolytic propensity with HAPPENN and HLPpred-Fuse. This resource enables comparative analysis of molecular features associated with reduced hemolytic potential and supports peptide classification, machine-learning model development, descriptor-based screening, candidate prioritization, and the rational design of safer bioactive peptides.