Deep learning discovers halicin, a structurally new broad-spectrum antibiotic
Result confirmed
Importance: major (4 of 5)The takeaway
MIT’s Collins and Barzilay labs (Cell, Feb 2020) trained a message-passing neural network on ~2,300 molecules.
Status
- Claim
Result confirmed
- Our reporting
- High confidence
- Verification
- Peer-reviewed
- Importance
- Major (4 of 5)
- Last verified
- 29 September 2026
Your AI and this story
- GPT-6 AstraIn its training data
- Claude Opus 5.5In its training data
- Gemini 4 HydrogenIn its training data
- Grok 4.7In its training data
It happened before the cutoffs of all four assistants, so it can be in their training data.
Key facts
- Published in Cell on 20 Feb 2020
- Screened >107 million molecules from ZINC15 in silico; of 23 top predictions tested, 8 were antibacterial
- Halicin treated C. difficile and pan-resistant A. baumannii infections in mice
- Structurally distant from known antibiotics; preclinical only
What happened
A graph neural network trained on growth-inhibition data screened the Drug Repurposing Hub and then 100M+ molecules, surfacing halicin and other candidates confirmed in the lab.
Why it matters
It launched the modern field of AI antibiotic discovery at a time of stalled antibiotic pipelines.
Sources
3 sources from 3 sites. Numbers match the chips in the text.
3 sources: 2 primary, 1 press
Primary
- A Deep Learning Approach to Antibiotic Discovery (Cell)cell.com, paper
- PubMed recordpubmed.ncbi.nlm.nih.gov, paper
Press
- Chemistry World: AI tool screens 107 million molecules, discovers potent new antibioticschemistryworld.com, press
Changes
- Filed