Optimal Descriptor Subset Search via Chemical Information and Target Activity-Guided Algorithm for Antimicrobial...

Area of Science:

• Computational chemistry and bioinformatics
• Drug discovery and development
• Peptide science

Background:

• Antimicrobial peptides (AMPs) show promise against multidrug-resistant pathogens.
• Computational AMP prediction methods exist, but shallow learning models rely heavily on manual feature engineering.
• Manual feature engineering may overlook critical information for AMP modeling.

Purpose of the Study:

• To investigate the application of the AExOp-DCS algorithm for optimizing descriptor subsets in AMP modeling.
• To evaluate if AExOp-DCS can identify descriptor sets that improve AMP prediction performance.
• To develop a more efficient computational approach for AMP discovery.

Main Methods:

• Utilized the AExOp-DCS algorithm for automatic feature domain optimization.
• Identified optimal descriptor subsets based on chemical structure and biological activity.
• Developed and evaluated AMP models using AExOp-DCS optimized descriptors.
• Compared performance against state-of-the-art AMP prediction models.

Main Results:

• AExOp-DCS identified descriptors with information content comparable to existing top models.
• Optimized descriptors demonstrated higher discriminative capacity.
• AMP models using AExOp-DCS descriptors achieved comparable performance with fewer features.
• The approach enables dimensionality reduction without sacrificing accuracy.

Conclusions:

• AExOp-DCS provides an efficient method for selecting optimal descriptors for AMP modeling.
• This approach enhances the efficiency of computational pipelines for AMP discovery.
• The freely available Java software AExOp-DCS-SEQ facilitates peptide descriptor search and AMP classification.