Relative Binding Free Energy Methods Based on the Alchemical Transformation Framework: An Effective Strategy for Predicting the Umami Recognition Threshold of Umami Peptide Derivatives

Abstract

Umami peptides integrating both flavor and nutritional properties possess numerous derivatives that pose challenges for prediction using absolute free energy prediction tools and peptide classification models. While relative binding free energy (RBFE) methods based on the alchemical transformation framework have demonstrated excellent performance in drug activity prediction, their application in taste activity prediction remains unexplored. Through scientific literature analysis, 611 umami peptides and derivatives were systematically organized and cataloged with their structures and recognition thresholds then clustered into eight clusters. AFEMs were employed to calculate the RBFE between the central structure of each cluster and its derivatives, followed by logarithmic fitting of the ratio (base 10) between these energies and their thresholds. RBFE achieved near-perfect accuracy in qualitative judgment and significantly outperformed approximation methods (docking, MM-GBSA, and MM-PBSA) in qualitative prediction (R² = 0.912). Dynamic equilibrium analysis identified conserved contacts, including HdB_147_ASP, HI_HdB_218_ASP, HI_HdB_220_TYR, HdB_SB_277_ARG, and HdB_301_GLU. Density functional theory elucidated how subtle structural differences alter electrostatic surfaces and Δ_HOMO–LUMO gaps, thereby driving threshold variations. The curated data set has been published as TPDB-Derivative (http://www.tastepeptides-meta.com/TastePeptides_Derivative). This study pioneered the application of AFEMs for activity prediction in umami research and provided technical foundations for the rational design of future umami peptides.