AMPActiPred: A three-stage framework for predicting antibacterial peptides and activity levels with deep forest

Area of Science:

• Computational biology
• Antimicrobial drug discovery
• Bioinformatics

Background:

• Antibiotic resistance is a major global health concern, driving the need for alternative antibacterial strategies.
• Antimicrobial peptides (AMPs), including antibacterial peptides (ABPs), show promise as novel therapeutic agents against bacterial infections.

Purpose of the Study:

• To develop AMPActiPred, a computational framework for identifying ABPs, characterizing their activity against various bacterial species, and predicting their activity levels.
• To provide a user-friendly web interface for accessing the AMPActiPred framework.

Main Methods:

• Utilized a three-stage computational framework incorporating peptide descriptors for compositional and physicochemical properties.
• Employed deep forest architecture for enhanced feature processing and predictive performance.
• Evaluated performance using metrics such as Accuracy, MCC, GMean, and PCC.

Main Results:

• Achieved state-of-the-art performance in ABP identification (Accuracy: 87.6%, MCC: 0.742).
• Demonstrated balanced prediction of ABP activity across 10 bacterial species (average GMean: 82.8%).
• Showcased robust prediction of ABP activity levels (average PCC: 0.722) and excellent interpretability.

Conclusions:

• AMPActiPred is the first computational framework capable of predicting both the targets and activity levels of ABPs.
• The framework offers a valuable tool for accelerating the discovery and development of novel antibacterial peptide therapeutics.
• A web interface is available at https://awi.cuhk.edu.cn/∼AMPActiPred/ for broader accessibility.