Researchers reported on Mar. 23 that machine learning has been used to speed up the discovery of antimicrobial peptides (AMPs) with potential as new treatments for ulcerative colitis, a chronic inflammatory bowel disease. The study, published in eGastroenterology by Miao and colleagues, combined computational screening with laboratory testing to identify promising AMP candidates.
Ulcerative colitis is marked by recurring intestinal inflammation, abdominal pain, and diarrhea. While current therapies such as 5-aminosalicylic acid, antibiotics, and biologics can help control symptoms, many patients experience incomplete responses or side effects. The need for safer and more effective treatments remains significant.
In this research, scientists developed a machine-learning pipeline that analyzed structural and chemical properties of over 6,000 peptide sequences. This approach identified 22 potential AMPs. Five were synthesized for experimental testing; one peptide named LR showed the best balance between antibacterial activity and low toxicity. Laboratory tests demonstrated LR’s strong ability to kill harmful bacteria like Escherichia coli and Staphylococcus aureus while maintaining good biocompatibility.
Testing in a mouse model of colitis revealed that LR treatment led to significant improvements in disease severity compared with both standard anti-inflammatory drugs and antibiotics. Mice treated with LR had less weight loss, lower disease activity scores, reduced colon shortening, improved tissue health under microscopic examination, lower levels of pro-inflammatory cytokines such as TNF-α and IL-6, and better intestinal barrier function indicated by increased tight junction proteins.
The study also found that LR influenced gut microbiota composition by increasing the abundance of Akkermansia muciniphila—a bacterium linked to improved gut health—while selectively inhibiting pathogenic bacteria without harming beneficial microbes. Supplementation with A. muciniphila alone partially alleviated colitis symptoms in mice.
Researchers concluded that integrating machine learning into drug discovery could streamline identification of new therapeutics for complex diseases like ulcerative colitis. They noted further studies are needed before translation into human treatment but highlighted the promise of microbiota-friendly approaches using artificial intelligence.
