Efficient Remote Homology Detection with Secondary Structure

Motivation: The function of an unknown biological sequence can often be accurately inferred if we are able to map this unknown sequence to its corresponding homologous family. Currently, discriminative approach which combines support vector machine and sequence similarity is recognized as the most accurate approach. SVM-Fisher and SVM-pairwise methods are two representatives of this approach, and SVM-pairwise is the most accurate method. However, these methods only encode sequence information into their feature vectors and ignore the structure information. In addition, one of their major drawbacks is their computation inefficiency. Based on this observation, we present an alternative method for SVMbased protein classification. Our method, SVM-I-sites, uses structure similarity instead of sequence similarity for remote homology detection. Our studies show that SVM-I-sites is much more efficient than both SVM-Fisher and SVMpairwise while achieving a comparable performance with SVM-pairwise. Result: We adopt SCOP 1.53 as our dataset. The result shows that SVM-I-sites runs much faster and is able to outperform many state-of-the-art sequence-based methods such as PSI-BLAST, SAM and SVM-Fisher, and comparable to SVM-pairwise. Availability: I-sites server is accessible through the web at http://www.bioinfo.rpi.edu. Programs are available upon request for academics. Licensing agreements are available for commercial interests. The framework of encoding local structure into feature vector is available upon request. Contact:[email protected], [email protected]