T1SEstacker: A Tri-Layer Stacking Model Effectively Predicts Bacterial Type 1 Secreted Proteins Based on C-Terminal Non-repeats-in-Toxin-Motif Sequence Features

Abstract

Type 1 secretion systems play important roles in pathogenicity of Gram-negative bacteria. However, the substrate secretion mechanism remains largely unknown. In this research, we observed the sequence features of repeats-in-toxin (RTX) proteins, a major class of type 1 secreted effectors (T1SEs). We found striking non-RTX-motif amino acid composition patterns at the C termini, most typically exemplified by the enriched “[FLI][VAI]” at the most C-terminal two positions. Machine-learning models, including deep-learning ones, were trained using these sequence-based non-RTX-motif features and further combined into a tri-layer stacking model, T1SEstacker, which predicted the RTX proteins accurately, with a fivefold cross-validated sensitivity of ∼0.89 at the specificity of ∼0.94. Besides substrates with RTX motifs, T1SEstacker can also well distinguish non-RTX-motif T1SEs, further suggesting their potential existence of common secretion signals. T1SEstacker was applied to predict T1SEs from the genomes of representative Salmonella strains, and we found that both the number and composition of T1SEs varied among strains. The number of T1SEs is estimated to reach 100 or more in each strain, much larger than what we expected. In summary, we made comprehensive sequence analysis on the type 1 secreted RTX proteins, identified common sequence-based features at the C termini, and developed a stacking model that can predict type 1 secreted proteins accurately.

DeepSecE: A Deep-Learning-Based Framework for Multiclass Prediction of Secreted Proteins in Gram-Negative Bacteria

“…No published multiclass prediction tool has considered all 5 major types (I to IV and VI) of secreted substrates of Gram-negative bacteria. Here, we selected 7 state-of-the-art binary predictors (T1SEstacker [ 15 ] for T1SE; BEAN 2.0 [ 17 ], Bastion3 [ 18 ], T3SEpp [ 19 ], and EP3 [ 21 ] for T3SE; Bastion4 [ 27 ], CNN-T4SE [ 28 ], iT4SE-EP [ 30 ], and T4SEfinder [ 31 ] for T4SE; and Bastion6 [ 32 ] for T6SE) for performance comparison. We used DeepSecE to predict the protein sequences in independent datasets and recorded the classification metrics.…”

“…In silico methods, particularly machine-learning-based approaches, have been developed and used to identify and distinguish secreted substrates from nonsecreted proteins [ 3 , 13 ]. For instance, some tools identify different types of secreted proteins, including type I secreted proteins (T1SEs) [ 14 , 15 ], T3SEs [ 16 – 22 ], T4SEs [ 23 – 31 ], and T6SEs [ 32 , 33 ]. Position-specific scoring matrix (PSSM)-based methods, such as Bastion3 [ 18 ], convolutional neural network (CNN)-T4SE [ 28 ], and Bastion6 [ 32 ], usually achieve a sound predictive performance.…”

Title: Toleration of Frameshift Mutations in mRNA Sequences Encoding the N-terminal Peptides of Bacterial Type III Effectors

“…The raw and frameshifting peptide sequences were predicted with T4SEpre (Wang et al, 2014), T1SEstacker (Chen et al, 2022) and SignalP 6.0 (Teufel et al, 2022) respectively.…”

“…The C-terminal 100-aa peptides were retrieved from the raw and frameshift-mutated type IV effectors, and predicted by T4SEpre with a cutoff of 3/3, that is, positive prediction by all the three modules, T4SEpre_bpbaac, T4SEpre_pseaac and T4SEpre_joint (Wang et al, 2014). T1SEstacker was used to predict the C-terminal 60-aa fragments of the raw and frameshift-mutated type I secreted proteins, with a default cutoff (>=3/6) (Chen et al, 2022). For the Sec/Tat signal peptides, we used SignalP 6.0 and the default settings to screen all possible Sec/Tat substrates from the raw and frameshift-mutated proteome of E. coli MG1655, S. typhimurium 14028S, Y. entericolitica 8081 and R. solanacearum GMI1000 (Teufel et al, 2022).…”