Backed by $1 billion, Xaira Therapeutics is readying AI-generated drugs

When Hetu Kamisetty was a postdoctoral researcher in David Baker’s lab at the University of Washington just over a decade ago, artificial intelligence was a nascent technology. Kamisetty built some early software models that could predict protein structures and then design them, but it was “the basic science era,” he says. “It wasn’t at that inflection point.”

That’s changed. Baker and his peers have begun using large language models and diffusion to dream up new proteins, generating different molecules just as ChatGPT generates text. And for the last eight months, Kamisetty has been leading artificial intelligence and machine learning efforts at a new start-up called Xaira Therapeutics, which formally launched Tuesday with $1 billion in venture capital—one of the largest venture deals in the history of the biotech industry. Endpoints News first reported the launch.

Xaira seeks to build on Baker’s RFdiffusion model, which is capable of designing, in just weeks, protein drugs that bind to influenza molecules in weeks(bioRxiv 2022, DOI: 10.1101/2022.12.09.519842). Xaira will also build on a more recent iteration called RFantibody, which binds to coronaviruses as well as influenza. Baker’s team published information on RFantibody to a preprint server just last month (bioRxiv 2024, DOI: 10.1101/2024.03.14.585103).

“The here and now . . . is our ability to make molecules with high specificity and control. This is literally the ‘generative’ part of the AI,” Kamisetty says. “The models have come a long way.”

Designing protein drugs from scratch would be a fundamental shift in the discovery and development of biologics, which currently involve screening blood for antibodies that might be suitable for known targets. The Xaira team hopes to reverse-engineer that process: take a validated target, then design an antibody that can bind to it well. Kamisetty expects to be able to apply the same process to small molecules down the line.

“The models understand protein chemistry very well at this point,” he says. “We think we are making progress towards understanding chemistry at its core, so we can generate matter of all kinds, not just protein matter.”

Xaira came to fruition last year in a stroke of lucky timing. Baker and Kamisetty had been discussing forming a start-up based on Baker’s diffusion technology, while the venture capitalists Bob Nelsen of Arch Venture Partners and Vik Bajaj of Foresite Capital were eyeing the rapidly growing artificial intelligence space. The three parties banded together, and Arch and Foresite quickly put their money toward the new venture. Other investors included F-Prime Capital, NEA, Sequoia Capital, Lux Capital, Lightspeed Venture Partners, Menlo Ventures, Two Sigma Ventures, and SV Angel, according to Endpoints.

Kamisetty leads Xaira’s site in Seattle. The company also has a site in South San Francisco, a small presence in the UK, and a smattering of remote employees. The idea is to hire where the talent is—the same reason Xaira is so well funded, Kamisetty says, since experts in artificial intelligence for drug making can be expensive.

Xaira has several big names attached to it. Several months ago, the start-up’s cofounders tapped Marc Tessier-Lavigne, a former Genentech scientist and onetime president of Stanford University, to lead Xaira as CEO. (Tessier-Lavigne stepped down as Stanford president last year after an investigation found significant flaws in papers he coauthored during his time at Genentech. Xaira’s press team declined to entertain questions about Tessier-Lavigne’s time at Stanford, but Tessier-Lavigne says he’s found his return to industry “energizing.”)

Xaira counts former US Food and Drug Administration commissionerScott Gottlieb; Stanford chemist, Nobel laureate, and recent Priestley Medalist Carolyn Bertozzi; and former Johnson & Johnson CEO Alex Gorsky among its board members, per Endpoints.

Altogether, Xaira employs about 60 people. Tessier-Lavigne declines to name any development timelines or elaborate on disease areas, instead saying the company will be “agnostic” in its search for new targets.

“The design technology will help improve antibody drug discovery in general, but we think we have a unique opportunity available today that wasn’t a year ago—or even six months ago,” Tessier-Lavigne says. “[We’re looking for] validated targets that are underserved by current technology.”