SherLoc2: A High-Accuracy Hybrid Method for Predicting Subcellular Localization of Proteins

摘要

SherLoc2 is a comprehensive high-accuracysubcellular localization prediction system. It is applicableto animal, fungal, and plant proteins and covers all maineukaryotic subcellular locations. SherLoc2 integrates sev-eral sequence-based features as well as text-based fea-tures. In addition, we incorporate phylogenetic profilesand Gene Ontology (GO) terms derived from the proteinsequence to considerably improve the prediction perfor-mance. SherLoc2 achieves an overall classification ac-curacy of up to 93% in 5-fold cross-validation. A novelfeature, DiaLoc, allows users to manually provide theircurrent background knowledge by describing a proteinin a short abstract which is then used to improve theprediction. SherLoc2 is available both as a free Webservice and as a stand-alone version at http://www-bs.informatik.uni-tuebingen.de/Services/SherLoc2.