MIT neural network finds halicin, active in vitro and in mice

AI system

System: Directed-message-passing neural network (D-MPNN) for antibacterial activity prediction (Chemprop-style, MIT)
Model: A directed-message-passing neural network.
Release: Exact release not established.
Application developer: Massachusetts Institute of Technology (Collins and Barzilay laboratories) with the Broad Institute
Role: The model predicted antibacterial activity to select compounds for laboratory testing.
Configuration: Not assessed
Inputs and outputs: Not assessed
Autonomy: Not assessed
Workflow: A growth-inhibition training set of a few thousand molecules. A message-passing network ranking large chemical libraries. Wet-lab confirmation of predicted actives and animal-model testing.
Human oversight: Not assessed
Relation to AI: The neural network's prediction selected halicin for testing from compounds not previously recognized as antibiotics. Wet-lab experiments by the authors and later by independent groups confirmed the activity. The outcome so far is a laboratory and animal-model result.
AI contribution: The neural network's prediction selected halicin for testing from compounds not previously recognized as antibiotics. Wet-lab experiments by the authors and later by independent groups confirmed the activity. The outcome so far is a laboratory and animal-model result.

Outcome identifiers: Not assessed
Benefit: A deep-learning screen identified halicin, active in vitro and in mice and confirmed by other labs; no human outcome.
Category: Improved Knowledge Discovery
Status: Not assessed
Beneficiary groups: Society.
Beneficiaries: Not assessed
Reach: Not assessed
Distribution and access: Not assessed
Duration: Not assessed
Expected?: Yes
Costs and harms: Not assessed

Studies: Not assessed
Study design: Not assessed
Measurement: Not assessed
Setting and sample: Not assessed
Measures: Not assessed
Comparison: vehicle control and the first-line antibiotic metronidazole in a mouse Clostridioides difficile infection model
Results: Reference: vehicle or the antibiotic metronidazole (lower clearance rate). Result: sterilization of 3 out of 4 mice after 72 hr of treatment, and 4 out of 4 mice after 96 hr of treatment.
Additional value from AI: Not assessed
Limitations: Not assessed
Denominators: Not assessed
Materials for replication: Not assessed

Participants: Developer: Massachusetts Institute of Technology (Collins and Barzilay laboratories) with the Broad Institute. Operator: MIT / Broad Institute / McMaster University research teams.
Jurisdictions: Benefit: Cambridge, United States. Developer: United States. Operator: United States.
Ownership and control: Not assessed
Dependencies: Not assessed
Cross-border connections: Not assessed
Geopolitical implications: Not assessed

Conditions for replication: Predicted compounds must be available for synthesis or purchase. Activity must be confirmed experimentally. A verified real-world outcome would require clinical use.
Replication attempts: 1 follow-up entries. Benefit recurred: 1.
Later findings: 2 December 2021. An independent laboratory at KACST in Riyadh confirmed halicin's in vitro activity against gram-positive and gram-negative strains, including a multidrug-resistant A. baumannii isolate; the compound remains preclinical.
Availability: Not assessed
Institutional decision: Not assessed
Grounds: Not assessed
Action and result: Not assessed
Success criterion: Not assessed

Assessed benefits: Society.
Depth: Not assessed
Reasons: Society: The sources document in vitro and murine results and independent laboratory confirmation. No individually attributable real-world outcome such as a treated patient or an approved drug.
Observed or intended: Not assessed
Rubric: Not assessed
Assessor and date: Not assessed
Alternative assessments: Not assessed
Fill: Independently studied

Kind: Not assessed
Related entries: Not assessed
Field: Biomedicine / antibiotic discovery. Academic research.
Purpose: Built to predict molecules that inhibit bacterial growth and surface antibiotic candidates. Halicin was that purpose.
Dates: Date basis unassessed: 2020 (precision unassessed). Updated: 21 October 2026.
Stage: Not assessed
Version: SHA-256: 0e6e61d4447e…

Peer-reviewed study Developer

Stokes, J. M. et al. (2020). A Deep Learning Approach to Antibiotic Discovery. Cell 180, 688–702 (PMC8349178).
Peer-reviewed study Independent author

Booq, R. Y., Tawfik, E. A., Alfassam, H. A., Alfahad, A. J. and Alyamani, E. J. (2021). Assessment of the Antibacterial Efficacy of Halicin against Pathogenic Bacteria. Antibiotics 10(12), 1480 (PMC8698312).
Peer-reviewed study Independent author

Hussain, Z. et al. (2022). Study on antibacterial effect of halicin (SU3327) against Enterococcus faecalis and Enterococcus faecium. Pathogens and Disease 80(1), ftac037.