PSORT-B: A Web-based Tool for Bacterial Subcellular Localization Prediction
Jennifer L. Gardy1, Cory A. Spencer2, Fiona S.L. Brinkman
1jlgardy@sfu.ca, Dept. of Molecular Biology and Biochemistry, Simon Fraser University; 2cspencer@sprocket.org, Dept. of Molecular Biology and Biochemistry, Simon Fraser University
Computational prediction of protein subcellular localization represents an important step in the genome annotation and drug discovery processes. Several predictors designed to analyze eukaryotic sequences are available, however the few available methods for prokaryotic localization prediction lack both the breadth and the precision of their eukaryotic counterparts. PSORT-B, a web-based predictive tool, was designed to provide the bacterial research community with a comprehensive and accurate alternative to existing predictive methods. Using a series of analytical modules, each designed to identify a single sequence feature known to correlate with a specific localization site, PSORT-B is able to return a final prediction and an associated confidence value for each of the 5 localization sites present in Gram-negative bacteria. The program displayed 97% precision in 5-fold cross-validation testing. Issues that arose during PSORT-B’s development and testing, including the handling of proteins resident at multiple localization sites and the importance of returning a final prediction of “unknown” are discussed, and the current effort to develop a Gram-positive version of the program is described. Selected whole genome analysis is also presented. PSORT-B will be available online at http://www.psort.org.