Prediction of alpha-helices in proteins based on thermodynamic parameters from solution chemistry

Current studies on the protein folding problem are chiefly pursued by two approaches; one statistically investigates the three-dimensional structures of proteins and the other focuses on their thermodynamic properties in solution. It remains to be demonstrated that information from these two sources are consistent and complementary. Using two sets of thermodynamic parameters experimentally obtained from solution studies, namely the helix propensities derived from small model peptides and hydrophobicity based on solvent transfer free energies for amino acids, we predict alpha-helices in native proteins. The correlation coefficient of our predictions on 98 non-homologous proteins is 0.26, a value comparable to that of the statistical prediction of Chou & Fasman but less than that from the state of the art prediction (0.39) based on neural network methodology. Helix propensities derived from different experimental systems in aqueous solutions are highly consistent for predicting alpha-helices in proteins. The helix propensities and amphiphilicity in primary sequences make independent contributions to the occurrence of alpha-helices. We also show that, with appropriate models, statistical analyses of known three-dimensional structures can provide thermodynamically relevant quantities. These findings indicate that for practical helix prediction, various sets of propensities have reached reasonable consensus. The report also suggests integrating solution thermodynamic studies with protein structural analyses, and the role of supersecondary structures on helices in proteins.