Neural network prediction of 3(10)-helices in proteins

Secondary structure prediction from the primary sequence of a protein is fu ndamental to understanding its structure and folding properties. Although s everal prediction methodologies are in vogue, their performances are far fr om being completely satisfactory. Among these, non-linear neural networks h ave been shown to be relatively effective, especially for predicting beta - turns, where dominant interactions are local, arising from four sequence-co ntiguous residues. Most 3(10)-helices in proteins are also short, comprisin g of three sequence-contiguous residues and two capping residues. In order to understand the extent of local interactions in these 3(10)-helices, we h ave applied a neural network model with varying window size to predict 3(10 )-helices in proteins. We found the prediction accuracy of 3(10)-helices (s imilar to 14 %), as judged by the Matthew's Correlation Coefficient, to be less than that of beta -turns (similar to 20 %). The optimal window size fo r the prediction of 3(10)-helices was about 9 residues. The significance an d implications of these results in understanding the occurrence of 3(10)-he lices and preferences of amino acid residues in 3(10)-helices are discussed .