IQSPred-PLM: An Interpretable Quorum Sensing Peptides Prediction Model Based on Protein Language Model.

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Quorum sensing (QS) is a cell-to-cell communication mechanism regulating bacterial cooperative behaviors.
• Quorum sensing peptides (QSPs) are vital signaling molecules, particularly in Gram-positive bacteria, influencing functions like virulence and biofilm formation.
• Existing QSP prediction tools require performance and interpretability enhancements.

Purpose of the Study:

• To develop a novel and high-performing model for predicting QSPs.
• To improve the accuracy and interpretability of QSP identification.
• To leverage advanced deep learning techniques for bacterial signaling molecule prediction.

Main Methods:

• Integration of protein language models (PLMs), specifically ESM-2, for peptide sequence encoding.
• Application of a multi-scale residual convolutional neural network (MSRes-CNN) for feature extraction.
• Dynamic feature integration using an adaptive weight modulation (AWM) module followed by a fully connected network for classification.

Main Results:

• IQSPred-PLM achieved outstanding predictive performance on a benchmark dataset.
• Key performance metrics include 97.50% accuracy (ACC), 0.951 Matthews correlation coefficient (MCC), and 0.990 area under the ROC curve (AUC).
• Case studies and interpretability analyses validated the model's effectiveness.

Conclusions:

• IQSPred-PLM represents a significant advancement in QSP prediction accuracy and interpretability.
• The model's performance highlights the potential of integrating PLMs and CNNs for biological sequence analysis.
• This tool can aid in understanding bacterial communication and developing targeted interventions.