Potential Sciences bets on physics to speed drug design — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Potential Sciences launched on November 5 with seed funding in the tens of millions from SovAI, Twin Path Ventures and other investors, while its founders withheld the exact figure and valuation.
- Potential Sciences plans to build large synthetic chemistry datasets by simulating interactions from atoms through small molecules, amino acids and proteins, then train AI predictors to reject molecules that cannot work in the clinic.
- Protein Data Bank is the single public 3D-atomic-structure dataset underpinning models at a large proportion of AI drug-discovery labs, but Hatem Helal says it is too small to train systems that reliably accelerate candidate searches.
- Potential Sciences joined SovAI’s portfolio, whose roughly $800 million fund can give companies access to the UK’s national AI supercomputer; the investor says that scale would be difficult for a small startup.
- Alan Fersht says accurately calculating molecular interactions and structural fluctuations in cells would revolutionize drug development, while Peter Wolynes compares current protein engineering to “the Romans, not of the Renaissance.”
- Peter Wolynes says faster candidate identification does not remove the need to test safety and efficacy in mice, sick patients and normal people, calling that work the rate-limiting step in bringing drugs to market.
- Hatem Helal says Potential Sciences has not settled on revenue or whether to release its models freely, given potential misuse, and that tools designed to minimize toxicity can be flipped to maximize it.
Why it matters: For pharmaceutical companies, Hatem Helal says Potential Sciences’ approach would reduce experimentation costs enough to make pursuing rare-disease drugs economical, backed by a seed round in the tens of millions. The catch: testing in mice, sick patients and normal people remains the rate-limiting step, so faster candidate design does not remove the time and cost to market.
Ask SkimNews
