UK Startup Potential Sciences Uses Physics Simulations to Speed Drug Discovery

A new British startup is betting that artificial intelligence can dramatically shorten the slow, costly process of designing drugs. Potential Sciences, which launched on November 5, plans to build large chemical datasets by simulating interactions between atoms, small molecules, amino acids and proteins using basic physics principles. From that data, it aims to train AI models capable of identifying highly targeted drug candidates while bypassing expensive lab experimentation.

"We want molecules that are never going to work in the clinic to fail fast," founder and CEO Hatem Helal told WIRED. The company says it has seed funding in the tens of millions of dollars from SovAI, the UK government's AI venture capital fund, and Twin Path Ventures, among others. Founders declined to disclose the exact figure or valuation.

Helal argues that the Protein Data Bank, the public database most AI drug discovery labs rely on, is too small to train sufficiently advanced models. Potential Sciences instead intends to generate far larger synthetic datasets through simulation. "Think of drug discovery as a search problem," he said.

Academics see promise in the approach. Alan Fersht, an emeritus professor of chemistry at the University of Cambridge, said that if interactions between all molecules in a cell could be calculated accurately, "that would indeed revolutionize drug development." Peter Wolynes, a chemistry professor at Rice University, said he expects improvement but cautioned that protein engineering remains at an early stage.

SovAI, established this year with roughly $800 million at its disposal and the ability to offer access to the UK's national AI supercomputer, has backed the company. Joséphine Kant, SovAI's head of ventures, said the physics-based angle sets Potential Sciences apart. The startup still faces a crowded market, and faster candidate identification does not remove the cost and time of safety and efficacy testing. Helal also acknowledged the risk that tools designed to minimize toxicity could be misused, and said the company has not settled on a revenue model or whether to release its models freely.

Source: Wired