A new feature selection method for computational prediction of type III secreted effectors

The type III secretion system (T3SS) is a specialised protein delivery system that plays an important role in pathogenic bacteria. However, the secretion mechanism has not been fully understood yet. Especially, the identification of type III secreted effectors is a notoriously challenging problem which has attracted a lot of research interests in recent years. In this paper, we introduce a machine learning method using amino acid sequence features for predicting T3SEs. We use a topic model called HMM-LDA to select useful features, and conduct experiments on Pseudomonas syringae as well as some other bacterial genomes. The cross-validation results on P. syringae data set show an improved prediction accuracy with the reduced feature set. The experimental results on the test sets also demonstrate that the accuracy of the proposed method is comparable to or better than the accuracies achieved by other available T3SE prediction tools.