Co-evolution of alpha-helical transmembrane protein residues: large-scale variant profiling and complete mutational landscape of 2277 known PDB entries representing 504 unique human protein sequences

Abstract Membrane proteins play fundamental roles in cellular function, yet the evolutionary dynamics of their amino acid composition remain poorly understood. Our current study investigates the substitutional landscape and evolutionary patterns of hydrophilic and hydrophobic residues in membrane α-helical proteins, addressing a significant gap in our knowledge of protein evolution. We analyzed 2277 high-resolution protein structures from the RCSB Protein Data Bank corresponding to 458 unique PDB structures, 504 UniProt transmembrane entries and their AlphaMissense predicted mutational libraries including more than 5.8 million amino acid substitutions, focusing on known transmembrane α-helical proteins in Homo sapiens . Our analysis showed that the pathological outcome of the substitutions is diverse, as nonpolar to polar changes showed higher pathological scores in general. Notably, F Y substitutions showed significantly lower pathological scores. Our further analysis revealed a significant asymmetry in the evolutionary frequencies of polar and nonpolar amino acids. We identified key residue pairs driving this asymmetry, with F Y, A T, V T and A S co-evolution diverging from the expected negative correlations (Spearman’s rho > 0.20, p T substitution via an alanine intermediate and the G N substitution via a serine intermediate lower their statistical barrier, which would otherwise require two sequential base changes. We propose two e volutionary g ame t heory (EGT) based models to explain their diversification, with partial correlation analysis on residue frequencies in homolog sequences. These mathematical insights suggest a previously unrecognized evolutionary pressure, potentially linked to functional diversification, which could be targeted to combat drug resistance. Our results offer insights into membrane protein evolution and may inform improved methods for protein structure prediction and design.