Exploration of Peptide-GPCR Specificity Beyond Homology and Tertiary Structure: Computational Insights into...

Peptide-responsive G protein-coupled receptors (GPCRs) play pivotal roles in a wide variety of physiological regulatory systems in animals. Despite the substantial expansion of the GPCR and endogenous peptide repertoire identified through large-scale genomic and peptidomic studies, a significant number of receptors remain orphan, particularly those for non-homologous or species-specific peptides. Conventional deorphanization approaches based on sequence or ligand similarity, or receptor structure, are frequently ineffective. These current shortcomings in elucidation of peptide-GPCR interactions result in a persistent gap between sequence information and functional characterization. This review consolidates recent advances in computational approaches that address this challenge by focusing on peptide-GPCR pairs rather than on sequences or structures of peptides and receptors. Pair-centric machine learning frameworks, originally developed in the field of chemical genomics, enable systematic prediction of peptide-GPCR interactions without sequence similarity. Our originally developed machine learning system, PD-incorporated SVM, which incorporates peptide-specific descriptors into the frameworks, has enabled the experimentally validated identification of novel peptide-GPCR pairs in a tunicate and a nematode, and has predicted GPCRs for lineage-specific neuropeptides in ctenophore. In addition to interaction prediction, the present study explores the interaction determinant likelihood (IDL) framework, which offers a mechanistic interpretation of predicted interactions. IDL extracts residue-level determinants of ligand recognition directly from trained models, thereby enabling the identification of peptide features and receptor residues that collectively govern interaction specificity. The application of this approach to closely related GPCR paralogs demonstrates the emergence of distinct peptide specificities through limited, context-dependent residue substitutions, without significant alterations in overall receptor structure. Together, these findings highlight interaction determinants as a unifying framework linking deorphanization, molecular mechanism, and evolutionary diversification of peptide-GPCR signaling.