Bayesian Segmentation using Residue Proximity for Secondary Structure and Contact Prediction

Secondary structure, residue contacts and contact numbers play an important role in tertiary structure determination of proteins. In the recent past, mainly due to non local interactions, the Bayesian segmentation approach has been successfully used for secondary structure prediction. In this paper, the performance of the Bayesian segmentation approach has been enhanced by taking residue contacts into account. The three state prediction accuracy increased by 2% when residue contacts were taken into account. Due to the inherent flexibility the Bayesian segmentation approach has been extended to infer residue contacts and contact numbers with the same segmentations. The proposed method achieved Cor_R values greater than 0.70 for protein sequence 1a62 and 1aba