Incorporating Global Information into Secondary Structure Prediction with Hidden Markov Models of Protein Folds

Valentina Di Francesco, Philip McQueen, Jean Garnier, and Peter J. Munson

Here we propose an approach to include global structural information in the secondary structure prediction procedure based on hidden Markov models (HMMs) protein folds. We first identify the correct fold or topology of a protein by means of the HMMs of topology families of proteins. Then the most likely structural model for that protein is used to modify the sequence of secondary structure states previously obtained with a prediction algorithm. Our goal is to investigate the effect on the prediction accuracy of including global structural information in the secondary structure prediction scheme, by means of the HMMs. We find that when the HMM of the predicted topology of a protein is used to adjust the secondary structure sequence, predicted originally with the Quadratic-Logistic method, the crossvalidated prediction accuracy (Q3) improves by 3%. The topology is correctly predicted in 68% of the cases. We conclude that this HMM based approach is a promising tool for effectively incorporating global structural information in the secondary structure prediction scheme.


This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.