Deep mutational scanning and machine learning uncover antimicrobial peptide features driving membrane selectivity.

Area of Science:

• Biochemistry
• Molecular Biology
• Drug Discovery

Background:

• Antimicrobial peptides (AMPs) show promise as therapeutics but often exhibit toxicity due to indiscriminate membrane disruption.
• Understanding the sequence-structure-function relationships governing AMP membrane selectivity is crucial for therapeutic development.
• Previous methods lacked the capacity to explore sufficient peptide sequence variation.

Approach:

• Developed deep mutational surface localized antimicrobial display (dmSLAY), a high-throughput method to comprehensively map residue importance and flexibility across an AMP sequence.
• Applied dmSLAY to Protegrin-1, identifying thousands of variants influencing antibacterial activity.
• Utilized dmSLAY datasets with machine learning to analyze over 5.7 million Protegrin-1 variants.

Key Points:

• dmSLAY reveals critical positional residue information for AMPs.
• Specific sequence modifications, such as avoiding large aromatic residues and disulfide bonds, enhance bacterial selectivity.
• Machine learning models trained on dmSLAY data predict sequence variants driving specific membrane interactions.

Conclusions:

• dmSLAY provides an innovative, high-throughput platform for elucidating AMP sequence-structure-function relationships.
• Optimized Protegrin-1 variants demonstrate improved bacterial specificity.
• This approach can accelerate the design of novel, safer synthetic peptide-based drugs.