Roughly one third of all approved drugs act on a single family of proteins: G-protein coupled receptors, or GPCRs. They sit in the cell membrane and govern how the body senses and signals. And yet almost no approved antibodies target them. The proteins are slippery, hard to purify, and notoriously difficult to raise antibodies against. For decades the industry has treated them, along with ion channels and other transmembrane receptors, as a wall.
Antiverse, a company founded in 2017 and headquartered in Cardiff, Wales, was built to climb that wall. Its pitch is direct: use generative AI to design antibody sequences from scratch for exactly these difficult targets, then test them fast in a wet lab engineered around the same biology. The company describes the approach as "lab-in-the-loop" - the models design, the lab measures, and the measurements feed back into the models.
The result, Antiverse says, is a route from a target to an optimized antibody in about four months, against the one-to-three-year timelines common in traditional discovery. It is a claim the company backs with blinded partner studies and a growing list of programmes, though, like most techbio numbers, it is best read as a working figure rather than a guarantee.
What makes the story worth telling is not only the science. It is where and how it is being built: a small team, outside the usual biotech hubs of Boston and Cambridge, working with a Tokyo-listed pharma and a leading US disease foundation on some of the least tractable problems in the field.
