Umami_IP: An integrated model and explanatory analysis for quantitative prediction of Umami recognition threshold.

Umami peptides have gained considerable attention due to high nutritional value and distinct flavor. However, because of experimental complexity and time cost, the identification of umami peptides encounters obstacles. Based on the recognition threshold documented in TastePeptidesDB (http://tastepeptides-meta.com/TastePeptidesDB), this study utilized molecular descriptors, fingerprints, and docking to construct an ensemble model. A novel predictive framework, termed Umami_IP, was subsequently developed to predict umami peptide recognition thresholds. The R of Umami_IP in the training/test set was 84.95 %/84.46 %, and the explained variance was 86.82 %/84.47 %, respectively. Explanatory analysis and density functional theory were implemented to demonstrate a strong correlation between the electrostatic surface potential and the recognition threshold. This investigation identified electrostatic channels in proximity to the active pocket. Through multiple sequence alignment, key motifs (120S ∼ 169Y, 179 K ∼ 229 L) on T1R1 were discovered. This research provides valuable guidance and ranking for the screening and identification of umami candidate peptides. Umami_IP is freely accessible at http://tastepeptides-meta.com/Umami_IP.