High Throughput Meta-analysis of Antimicrobial Peptides for Characterizing Class Specific Therapeutic Candidates: An in-silico Approach

The increasing incidence of antimicrobial resistance (AMR) is becoming a serious concern worldwide and requires newer drugs. Recent evidence has shown growing interest in developing peptide-based therapeutics to tackle AMR. In the current study, we performed a meta-analysis of nearly 8.6 million predicted antimicrobial peptides (AMPs) and assessed their antibacterial, antiviral, and antifungal activity. We created high-quality, class-specific datasets and performed several analyses, including amino acid composition, motif preference, physiochemical properties, etc. We observed significant differences in the residue composition, charge, molecular weight, pI value, and instability index of peptides among 3 AMP classes. We further developed multiple machine learning models to predict peptide activity using the dataset. Our Extratree model developed using composition-based features achieved the highest AUROC of 0.98 for antibacterial peptide (ABP), 0.99 for antiviral peptide (AVP), and 0.99 for antifungal peptide (AFP) prediction when tested on an independent dataset. Subsequent filtering of peptides based on moonlighting properties (toxicity, allergenicity, cell-penetrating ability, half-life, and secondary structure) yielded a list of peptides that exhibit substantial therapeutic potential. We further selected the top 10 peptides in each category based on their half-lives, predicted their 3D structures using ColabFold, a feature built into ChimeraX1.8 software, and used HDock to perform molecular docking analysis with a pathogenic protein selected from an organism in each class. Docking studies demonstrated strong interaction between peptides and proteins, and based on free energy, we ranked the peptides. Overall, we have put forth class-specific peptides with high therapeutic potential based on rigorous meta-analysis encompassing [~]8.6 million AMPs.