Combining three scoring algorithms for representing protein sequence

Effective representation of the protein sequence is a key issue in detecting remote protein homology. Recent work using string kernels for protein data has achieved state-of-the-art performance for protein classification. However, such representations are suffering from high dimensionality problem. In this work, we introduce a simple method based on representing the protein sequence by fix dimensions of the length three. We present hidden Markov model combining scores method. Three scoring algorithms are combined to represent protein sequence of amino acids for better remote homology detection. We tested the method on the SCOP version 1.37 dataset. The results show that, with such a simple representation, we are able to achieve superior performance to previously presented protein homology detection methods while achieving better computational efficiency.