An antibiotic candidate discovered by a machine-learning model has entered its first human trials, a Phase 1 study enrolling 80 healthy volunteers to test its safety. The compound targets several bacterial strains that have grown resistant to existing drugs, a threat that contributes to an estimated 1.3 million deaths worldwide each year.
The model did not invent chemistry we could never have found. It searched a space too large for us to search by hand, and it searched it in weeks.
- Dr. Lena Fischer, infectious-disease researcher, Central Medical School
The research team trained the model to predict antibacterial activity, then let it screen more than 100 million candidate molecules. A shortlist was synthesised and tested in the laboratory, where the lead compound proved effective against three priority resistant strains.
Why resistance makes this urgent
The pipeline of genuinely new antibiotics has been thin for decades, in part because they are commercially unrewarding compared with drugs patients take for life. Computational discovery could lower the cost of finding new candidates, though each must still pass the same clinical trials as any other drug.
What Phase 1 will and will not show
The current study measures safety and dosing in healthy people, not whether the drug cures infections; that requires later, larger trials that will take years. Researchers stressed that a promising laboratory result is the beginning of a long road, not the end of one.