ROLE OF MAIN-CHAIN ELECTROSTATICS, HYDROPHOBIC EFFECT AND SIDE-CHAIN CONFORMATIONAL ENTROPY IN DETERMINING THE SECONDARY STRUCTURE OF PROTEINS

The physiochemical bases of amino acid preferences for alpha-helical, beta-strand, and other main-chain conformational states in proteins is con troversial. Hydrophobic effect, side-chain conformational entropy , steric factors, and main-chain electrostatic interactions have all b een advanced as the dominant physical factors which determine these pr eferences. Many attempts to resolve the controversy have focused on sm all model systems. The disadvantage of such systems is that the amino acids in small molecules are largerly exposed to the solvent. In prote ins, however, the amino acids are in contact with the solvent to a dif ferent degree, causing a large variability of strengths of all interac tions. The estimates of mean strengths of interactions in the actual p rotein environment are therefore essential to resolve the controversy. In this work the experimental protein structures are used to estimate the mean strengths of various interactions in proteins. The free ener gy contributions of the interactions are implemented into the Lifson-R oig theory to calculate the helix and strand free energy profiles. Fro m the profiles the secondary structures of proteins and peptides are p redicted using simple rules. The role of hydrophobic effect, side-chai n conformational entropy, and main-chain electrostatic interactions in determining the secondary structure of proteins is assessed from the abilities of different models, describing stability of secondary struc tures, to correctly predict alpha-helices, beta-strands and coil. in 1 30 proteins. The three-state accuracy of the model, which contains onl y the free energy terms due to the main-chain electrostatics with 40 c oefficients, is 68.7%. This accuracy is approaching to the accuracy of currently the best secondary structure prediction algorithm based on neural networks (72%); however, many thousands of parameters have to b e optimized during the training of the neural networks to reach this l evel of accuracy. The correlation coefficient between the calculated a nd the experimental helix contents of 37 alanine based peptides is 0.9 1. If the hydrophobic and the side-chain conformational entropy terms are included into the helix-coil transition parameters, the accuracy o f the algorithm does not improve significantly. However, if the, main- chain electrostatic interactions are excluded from the helix-coil and strand-coil transition parameters, the accuracy of the algorithm reach es only 59.5%. These results support the dominant role of the short-ra nge main-chain electrostatics in determining the secondary structure o f proteins and peptides. The role of the hydrophobic effect and the si de-chain conformational entropy is small. (C) 1998 Academic Press Limi ted.