Abstract:Many Gram-negative bacteria infect hosts and cause diseases by translocating a variety of type III secreted effectors (T3SEs) into the host cell cytoplasm. However, despite a dramatic increase in the number of available whole-genome sequences, it remains challenging for accurate prediction of T3SEs. Traditional prediction models have focused on atypical sequence features buried in the N-terminal peptides of T3SEs, but unfortunately, these models have had high false-positive rates. In this research, we integrated promoter information along with characteristic protein features for signal regions, chaperone-binding domains, and effector domains for T3SE prediction. Machine learning algorithms, including deep learning, were adopted to predict the atypical features mainly buried in signal sequences of T3SEs, followed by development of a voting-based ensemble model integrating the individual prediction results. We assembled this into a unified T3SE prediction pipeline, T3SEpp, which integrated the results of individual modules, resulting in high accuracy (i.e., ∼0.94) and >1-fold reduction in the false-positive rate compared to that of state-of-the-art software tools. The T3SEpp pipeline and sequence features observed here will facilitate the accurate identification of new T3SEs, with numerous benefits for future studies on host-pathogen interactions.IMPORTANCE Type III secreted effector (T3SE) prediction remains a big computational challenge. In practical applications, current software tools often suffer problems of high false-positive rates. One of the causal factors could be the relatively unitary type of biological features used for the design and training of the models. In this research, we made a comprehensive survey on the sequence-based features of T3SEs, including signal sequences, chaperone-binding domains, effector domains, and transcription factor binding promoter sites, and assembled a unified prediction pipeline integrating multi-aspect biological features within homology-based and multiple machine learning models. To our knowledge, we have compiled the most comprehensive biological sequence feature analysis for T3SEs in this research. The T3SEpp pipeline integrating the variety of features and assembling different models showed high accuracy, which should facilitate more accurate identification of T3SEs in new and existing bacterial whole-genome sequences.

TSignal: a transformer model for signal peptide prediction

Prediction of signal peptide cleavage sites with subsite-coupled and template matching fusion algorithm

SignalP 5.0 improves signal peptide predictions using deep neural networks

Signal-3L: A 3-layer approach for predicting signal peptides.

Unbiased organism-agnostic and highly sensitive signal peptide predictor with deep protein language model

T4SEpp: a pipeline integrated with protein language models effectively predicting bacterial type IV secreted effectors

Signal-CF: a subsite-coupled and window-fusing approach for predicting signal peptides.

Signal-3L 2.0: A Hierarchical Mixture Model for Enhancing Protein Signal Peptide Prediction by Incorporating Residue-Domain Cross-Level Features.

Signal-3L 3.0: Improving Signal Peptide Prediction Through Combining Attention Deep Learning with Window-Based Scoring

Prediction of Signal Peptide Cleavage Sites with Template Matching Fusion Algorithm

Predicting protein N-terminal signal peptides using position-specific amino acid propensities and conditional random fields

Identification and Analysis of the Cleavage Site in a Signal Peptide Using SMOTE, Dagging, and Feature Selection Methods.

Improve Signal Peptide Prediction by Using Functional Domain Information.

Signal Peptide Prediction Based on Analysis of Experimentally Verified Cleavage Sites.

Effects of Neighboring Sequence Environment in Predicting Cleavage Sites of Signal Peptides.

Enhance the Recognition of Signal Peptide Domain and Transmembrane Domain Based on Transformer

T3SEpp: an Integrated Prediction Pipeline for Bacterial Type III Secreted Effectors

Predicting Deleterious Non-Synonymous Single Nucleotide Polymorphisms in Signal Peptides Based on Hybrid Sequence Attributes

PepFormer: End-to-End Transformer-Based Siamese Network to Predict and Enhance Peptide Detectability Based on Sequence Only

Computational Prediction of the Functional Effects of Amino Acid Substitutions in Signal Peptides Using a Model-Based Approach.

TULIP — a Transformer based Unsupervised Language model for Interacting Peptides and T-cell receptors that generalizes to unseen epitopes