Abstract
Amino acid similarity search methods are fundamental tools for inferring protein structure and function. Optimization of amino acid substitution matrices is indispensable to improve search performance. We explored matrices that are ideal for detecting distantly related proteins in the PCA subspace that represents a set of typical existing matrices, and derived a novel sensitive matrix: MIQS. Test results show that the MIQS detection performance is superior to those of existing matrices and other sequence comparison methods. This refinement of the amino acid substitution matrix might influence many fields of sequence analysis. MIQS is available at http://csas.cbrc.jp/Ssearch/.
