Predicting quorum sensing peptides using stacked generalization ensemble with gradient boosting based feature...

Area of Science:

• Microbiology
• Computational Biology
• Bioinformatics

Background:

• Bacteria utilize quorum sensing (QS) molecules, primarily peptides in Gram-positive species, to regulate critical functions within multicellular aggregates.
• Targeting QS molecules with antibodies presents a promising therapeutic strategy for bacterial control.
• Efficient identification of QS peptides is essential for advancing high-throughput experimental research.

Purpose of the Study:

• To develop a fast, reliable, and accurate predictive model for identifying QS peptides.
• To enhance the efficiency of experimental identification of QS peptides for therapeutic and research applications.

Main Methods:

• Development of a stacked generalization ensemble model named EnsembleQS.
• Utilized Gradient Boosting Machine (GBM) for feature selection (791D features).
• Evaluated model performance against baseline classifiers and on an independent dataset.

Main Results:

• EnsembleQS demonstrated superior performance compared to baseline classifiers.
• Achieved an accuracy of 93.4% on an independent dataset of 40 QS peptides.
• Reported a Matthew's Correlation Coefficient (MCC) of 0.91 and an area under the ROC curve (AUC) of 0.951.

Conclusions:

• EnsembleQS is a highly accurate computational framework for predicting QS peptides.
• The model effectively supports proteomics research by facilitating QS peptide identification.
• The developed tool and datasets are publicly available to aid further research.