New drug classes, discovered at machine speed

AI proven in antibacterials, built to power discovery beyond infectious disease

Heterogeneous data is an asset, not noise

Scientists and ML practitioners often discard heterogeneous bioactivity data because absolute values don’t reproduce across assays. We treat that as a modeling opportunity, not a data problem.

The result: models proven in real drug discovery — on small datasets and novel chemical classes where conventional approaches fail.

ML-enabled discovery, in numbers

Lead optimization on our first antibacterial program ran roughly three times faster than typical, with about one-third the compounds and one-third the team

Stat-card comparison: typical antibacterial program — 4–5 years lead optimization, 1,200–2,500 compounds made, 20–25 person team — versus Project 1 ML-enabled — 17 months lead optimization, 697 compounds made, 8 person team.

Quarterly bar chart of best-in-series percent oral bioavailability for the lead program from 3Q23 through 1Q26 — bars stay at or below 22% through 2Q25, then jump to roughly 64% in 3Q25, 84% in 4Q25, and 89% in 1Q26 after the oral bioavailability model was deployed.

Solving oral bioavailability — one of medicinal chemistry’s hardest problems

Model trained on public data alone, never seeing our chemistry — generalizable to small molecules from any therapeutic area
Run on every molecule made up to deployment, it identified the progenitor of our lead series — driving us to >80% bioavailability from a series we otherwise would have deprioritized

Proving the platform in antibacterials
Built to generalize beyond

Our lead program is a novel oral small molecule for outpatient urinary tract infections, on track for pre-DC nomination by mid-2026. A second oral program targeting non-tuberculous mycobacterial (NTM) infections is progressing rapidly behind it.

The data treatment, the models, and the indication-first workflow weren’t built for one disease area — they generalize to any therapeutic area.

Strategy & Pipeline →

Get in touch

We welcome partnering and investor conversations about the platform and our programs.

References

Smith P.A., et al. Optimized arylomycins are a new class of Gram-negative antibiotics. Nature 561, 189–194 (2018). doi:10.1038/s41586-018-0483-6
Basarab G.S., et al. Responding to the challenge of untreatable gonorrhea: ETX0914, a first-in-class agent with a distinct mechanism-of-action against bacterial Type II topoisomerases. J. Med. Chem. 58, 6264–6282 (2015). doi:10.1021/acs.jmedchem.5b00863