The Challenge

Antimicrobial resistance threatens global health, and designing antimicrobial peptides requires balancing efficacy and toxicity. Madam Therapeutics needed to predict activity and toxicity of synthetic peptides quickly to accelerate discovery.

Our Approach

We combined large‑scale data mining, machine learning and insight extraction to accelerate antimicrobial peptide discovery.

• Data mining & database curation: Used NLP and text mining to compile the largest annotated AMP database, capturing sequences, activity and toxicity information.
• Machine learning modelling: Built and validated predictive models for antimicrobial activity and toxicity across multiple bacterial species.
• Insight extraction & improvement: Explored correlations between toxicity types and physicochemical properties, and curated the database for ongoing improvements.

The Outcome

• Created the largest annotated AMP database to date.
• Demonstrated models with precision and recall values as high as 0.74 and 0.96 respectively.
• Identified promising compounds now progressing into clinical development.
• Enabled data‑driven design of new antimicrobial peptides.

Why It Matters

By combining NLP and ML, this project turned peptide discovery into a data‑driven process. Predictive modelling reduces experimental screening time and supports development of safer antimicrobials in the fight against antimicrobial resistance.