Computational Design of Potentially Multifunctional Antimicrobial Peptide Candidates via a Hybrid Generative Model

Area of Science:

• Biotechnology
• Computational Biology
• Infectious Diseases

Background:

• Antimicrobial peptides (AMPs) are crucial alternatives to conventional antibiotics due to their broad-spectrum activity and ability to overcome microbial resistance.
• Current deep learning methods for AMP generation struggle with creating multifunctional peptides due to complex amino acid interactions and limited functional activity prediction.

Purpose of the Study:

• To develop a novel de novo multifunctional AMP design framework to address limitations in current deep learning approaches.
• To enhance the generation of AMPs with multiple, optimized antimicrobial functionalities.

Main Methods:

• Integration of a Feedback Generative Adversarial Network (FBGAN) with a global quantitative AMP activity regression module and a multifunctional-attribute integrated prediction module.
• Utilizing pre-trained regression and classification models with feedback loops for predicting Minimum Inhibitory Concentration (MIC) values.
• Employing a combinatorial predictor to simultaneously identify and predict five multifunctional AMP bioactivities.

Main Results:

• The framework successfully facilitates automated generation of potential AMP candidates with enhanced multifunctionality.
• Optimized computational predictions for AMP activity, including MIC values.
• Demonstrated efficiency in generating AMPs with multiple, enhanced antimicrobial properties.

Conclusions:

• The developed framework provides an effective approach for designing multifunctional AMPs.
• This work offers a valuable reference for developing novel strategies against multi-drug-resistant infections.
• The integrated deep learning model enhances the prediction and generation of AMPs with diverse bioactivities.