🧬Potential Sciences Raises Tens of Millions to Simulate Drug Discovery With AI

Drug discovery might stop depending on the lab bench

A British startup called Potential Sciences wants to skip most of the lab work in drug discovery. The plan is to simulate interactions between atoms, small molecules, amino acids, and proteins, then use that synthetic data to train AI models that pick out drug candidates before anyone touches a pipette. Why should you care? If you build anything touching bioinformatics, or you just like watching ML hit a hard real-world problem, this is a useful test case. The real bottleneck in drug design is data, and the public Protein Data Bank is still too small to train models that reliably find viable candidates. Synthetic data is the workaround, and the question is whether it holds up. The numbers are thin on purpose. The founders declined to share the seed round size or valuation, and the company only says it is in the tens of millions, backed by SovAI and Twin Path Ventures. SovAI itself was set up earlier this year by the UK government with roughly $800 million and access to the national AI supercomputer. The catch? Simulated binding predictions have to match real-world results, and nobody has proven that yet. Potential Sciences joins Basecamp Research, Isomorphic Labs, and CuspAI in this space, so expect comparisons soon.