A comparative study on sequence feature extraction for type III secreted effector prediction

Protein secretion is an essential mechanism for bacterial survival in their surrounding environment. The type III secretion system (T3SS) is a specialized protein delivery system that plays a key role in pathogens. Since the secretion mechanism has not been fully understood yet, T3SS has attracted a great deal of research interests. Especially, the identification of novel effectors (secreted proteins) is an important and challenging task for the T3SS study. This paper adopts machine learning methods to predict type III secreted effectors (T3SE). We conduct a comparative study on the feature extraction methods for protein sequence of T3SEs, and propose new methods involving sequence features, secondary structure and solvent accessibility information. The experimental results on Pseudomonas syringae data set demonstrate the effectiveness of our methods.