AI-Driven Breakthrough: First New Antibiotics in Over Six Decades

A team of 21 MIT and Harvard researchers used deep-learning to screen ~12 million compounds for activity against methicillin-resistant *Staphylococcus aureus* (MRSA). Testing ~280 molecules identified two candidates that cut MRSA levels tenfold in mouse skin and systemic infection models.

Euronews

Antibiotic resistance is a crisis. The European Centre for Disease Prevention and Control (ECDC) reports ~150 000 MRSA infections and 35 000 antimicrobial-resistant deaths annually in the EU. No new antibiotic class has entered the market in the past 60 years, underscoring the need for therapies.

Euronews

James Collins (MIT professor, senior author), Felix Wong (MIT & Harvard postdoc, lead author), and their MIT-Harvard team (21 scientists) performed model development and validation. ECDC supplied epidemiological data; the findings were published in *Nature*.

Euronews

• Training: ~39 000 compounds were evaluated for MRSA activity, forming the basis of an enlarged deep-learning model. Three auxiliary models predicted toxicity on human cells.

Euronews

• Screening: Integrated predictions narrowed 12 million compounds to five chemical subclasses; ~280 were purchased for testing. Two compounds showed strong in-vitro activity and achieved a tenfold bacterial reduction in mouse models.

Euronews

These are the first new antibiotic candidates in more than six decades, offering treatment for MRSA-driven skin, pneumonia, and bloodstream infections. If clinical trials succeed, they could blunt the projected rise in drug-resistant infections and related deaths.

Euronews

Researchers emphasized the AI framework’s speed and its ability to provide mechanistic insight. They described the work as an effort to “open the black box” of deep-learning predictions, aiming for a more transparent drug-discovery process.

Euronews

No dissenting viewpoints or external criticism were reported.

Euronews

Results are limited to preclinical mouse studies; human efficacy, safety, and pharmacokinetics remain untested. Chemical structures of the lead compounds and their regulatory pathways were not disclosed.

Euronews

• “The insight here was that we could see what was being learned by the models to make their predictions that certain molecules would make for good antibiotics,” — James Collins, MIT

Euronews

• “Our work provides a framework that is time-efficient, resource-efficient, and mechanistically insightful, from a chemical-structure standpoint, in ways that we haven’t had to date” — James Collins, MIT

Euronews

• “What we set out to do in this study was to open the black box. These models consist of very large numbers of calculations that mimic neural connections, and no one really knows what's going on underneath the hood,” — Felix Wong, MIT & Harvard

Euronews

The team will pursue further preclinical safety studies and, pending regulatory clearance, launch Phase I trials. Ongoing AI refinements aim to speed discovery of antimicrobial agents.

Euronews