Importance of Secondary Structure Elements for Prediction of Go Annotations

Predicted or actual protein secondary structure, in addition to amino acid sequence, is often used for fold recognition and function prediction. Different kinds of secondary structure elements could be predicted with different accuracy by different prediction methods and this could affect the fold or function prediction performance. In this study, contribution of amino acid sequence residues belonging to different types of secondary structure elements (H: alpha helix, E: beta sheet, L: loop) for protein function prediction is investigated. Smith-Waterman alignment similarity scores between amino acid sequences belonging to 6 different sets of secondary structure elements, namely, HEL, HE, HL, H, E and L, are computed. Using these alignment scores, protein function prediction is performed. On a function prediction data set, consisting of 27 Gene Ontology (GO) classes and 4498 sequences, it is found out that using the whole amino acid sequence results in the best performance. Using H and L regions together results almost as well performance as HEL. E regions alone are the least significant in function prediction.