A deep learning framework for sequence-based bacteria type IV secreted effectors prediction

摘要

Type IV secretion system (T4SS) is a specialized protein delivery system in gram-negative bacteria that injects proteins (called effectors, T4SEs) directly into the eukaryotic host cytosol and facilitates bacterial infection. Since various T4SEs have been experimentally validated to play important roles in a wide variety of biological activities, identifying them is crucial to our understanding of host-pathogen interactions and bacterial pathogenesis. However, experimental identification is often time-consuming and expensive. In the post-genomic era, it becomes imperative to predict new T4SEs using information from the amino acid sequence alone when new proteins are being identified in a high-throughput mode. Consequently, in this work we propose, DeepT4, a novel deep learning method to directly classify any protein sequence into T4SEs or non-T4SEs only using the protein primary sequences. The backbone of our framework is a convolutional neural network (CNN), which automatically extracts T4SEs-related features from 50 N-terminal and 100 C-terminal residues of the protein. We train and test the deep CNN model on a comprehensive dataset across multiple bacterial species, with a high receiver operating curve of 0.876 in the 5-fold cross validation and an accuracy of 92.2% for the test set. Moreover, when performing on a common independent dataset, DeepT4 outperforms known sequence-based state-of-the-art T4SEs prediction methods. We believe that deep learning is a valuable method to predict type IV secreted effectors. This study will be useful in elucidating the secretion mechanism of T4SS and facilitating hypothesis-driven experimental design and validation.