A Hidden Markov Model approach to model protein sequence and structural information: Identification of helix-turn-helix DNA-binding motif

This study presents a Hidden Markov Model (HMM) approach to model protein sequence and structure-derived information. The HMM emits both amino acid residue identity and the solvent accessibility of residues. The solvent accessibility of each residue is discretized into three states: buried (B), mediate (M) and exposed (E). A set of standard helix-turn-helix (HTH) motifs from a set of heterogeneous DNA-binding proteins was used to develop the HMM model for HTH motifs. The resulting HMM can identify HTH with a higher confidence and a higher sensitivity than the HMM that models only sequence information. It can also identify HTH motifs with structures different from the standard HTH motif.