2004 Volume 4 Issue 4 Pages 142-147
We describe a novel method for predicting a signal peptide of which three-domain (tripartite) structure is recognized by three modules of the software system. The first module numerates hydrophobic segment in N-terminal 100 residues, the second predicts signal sequences including both signal peptides and signal anchors, and the third discriminates signal peptides. Two novel indexes, SS- and SP-indexes, were developed for the discrimination of signal sequences and signal peptides, respectively, by calculating the relative propensities of amino acids at the carboxyl-terminal end of the hydrophobic region. The number of adjustable parameters in the whole system was only five. When three groups of data (917 signal peptides, 103 signal anchors and 544 non-signal sequences) were analyzed, signal peptides of eukaryotes could be discriminated with the Matthews correlation coefficient of 0.89. The signal peptide predictor SOSUIsignal is available at the web site: http://bp.nuap.nagoya-u.ac.jp/sosui/sosuisignal/sosuisignal_submit.html. This system has the advantage of very fast calculation.