Ensemble learning based on bi-directional gated recurrent unit and convolutional neural network with word embedding...

Area of Science:

• Biochemistry
• Computational Biology
• Bioinformatics

Background:

• Bioactive peptides are crucial for various physiological functions, including antimicrobial and anticancer activities.
• Current methods for peptide classification and activity prediction are costly and time-consuming.
• There is a need for efficient computational tools to accelerate the identification and development of bioactive peptides.

Purpose of the Study:

• To develop and validate a novel computational model, BioPepPred-DLEmb, for predicting bioactive peptide activities.
• To enhance the accuracy and efficiency of peptide classification and function prediction.
• To identify key biomarkers for differentiating peptide activity states.

Main Methods:

• Integration of Convolutional Neural Networks (CNNs) and Bidirectional Gated Recurrent Units (BiGRUs).
• Application of natural language processing for encoding amino acids into dense vectors.
• Evaluation across nine diverse bioactive peptide datasets.
• Utilizing UMAP visualization and Kplogo analysis for biomarker identification.

Main Results:

• BioPepPred-DLEmb achieved superior predictive accuracy (0.909) and sensitivity (0.911) compared to traditional methods.
• The model effectively differentiated peptide activity states and identified key biomarkers.
• Predicted antimicrobial peptides (Pred-AMPs) demonstrated potent in vitro efficacy against Escherichia coli and Acinetobacter baumannii at low micromolar concentrations (2-16 µmol/L).

Conclusions:

• BioPepPred-DLEmb offers a robust and accurate computational approach for bioactive peptide development.
• The model facilitates advancements in precision medicine, personalized therapies, and functional food innovation.
• This tool accelerates the discovery pipeline for novel antimicrobial peptides with significant therapeutic potential.