Predicting Gram-positive bacterial protein subcellular localization based on localization motifs

The subcellular localization of proteins is closely related to their functions. In this work, we propose a novel approach based on localization motifs to improve the accuracy of predicting subcellular localization of Gram-positive bacterial proteins. Our approach performed well on a five-fold cross validation with an overall success rate of 89.5%. Besides, the overall success rate of an independent testing dataset was 97.7%. Moreover, our approach was tested using a new experimentally-determined set of Gram-positive bacteria proteins and achieved an overall success rate of 96.3%.

► We introduce a novel approach based on localization motifs to improve the predicting accuracy.
► We introduce a novel position-specific frequency encoding approach.
► Our approach greatly improved the accuracy of prediction.