In one of their first ever acquisitions, Relay bets $85M cash on a new AI-based screening approach

April 29, 2021

From: Endpoints News

Although they’ve never been short for cash, Relay Therapeutics hasn’t been one for acquisitions in its 5-year history, focusing instead on developing its own tools to study how proteins move and advancing molecules off those insights.

On Friday, though, the Third Rock-spun biotech plunked down $85 million in cash and another $185 million in milestones to acquire the small, two-year-old, Google-partnered machine learning company ZebiAI. The deal will allow Relay to add a critical new technology to its early-stage discovery tools now that, with three candidates in the clinic, they’ve shown those tools can pay off, said CEO Sanjiv Patel.

“It makes the whole discovery process much more efficient and effective,” Patel told Endpoints News of ZebiAI’s platform.

ZebiAI is one of a series of biotechs that have sprouted up over the last few years promising to discover drugs faster with a new technology called DNA-encoded library screens. By attaching different DNA strands to millions or billions of molecules, they can blast those molecules at a particular protein target and, with DNA sequencing, figure out which molecules hit and which didn’t.

Patel said that ZebiAI stood out to him from the other companies in the space after he read a paper they published in the Journal of Medicinal Chemistry last summer. For one, they had actually shown in a prestigious journal their technology could pay off. But they had also taken a slightly different tack than their rivals.

Founded by Richard Wagner, who also built the well-partnered biotech X-Chem, ZebiAI tried to solve some of the problems raised by DNA encoded libraries: Plenty of molecules will hit a given target but many may not be remotely suitable as drugs; you need to figure out which to pursue. ZebiAI developed a machine learning model to predict which molecules those would be. They also relied on commercial libraries of molecules that already have drug-like properties, weeding out ill-fitting molecules from the start.

“The goal is to get to chemical starting points that look like drugs much more rapidly,” Patel said. “And you only synthesize something in the wet lab when you get very close to essentially what could be a drug-like molecule.”

Relay will incorporate the ZebiAI tech into a platform that has attracted considerable buzz — and considerable cash — over the last half-decade and will see its first human data later this year. Founded on Brandeis University biochemist Dorothee Kern’s idea that you could find better drugs by using new technology to “look” at proteins in motion rather than as static objects, they’ve now brought three cancer drugs into the clinic and have an undisclosed pipeline of molecules for genetic diseases on the way.

Relay will principally use the screens to hunt for molecules that can hit the handful of proteins they’re already been working on. But, through a pre-existing initiative from ZebiAI, academic labs — which are often in sore need of chemical matter — will continue to use the platform to screen for molecules against new targets.

Relay may look to license or partner on programs that come out of the initiative.

“It’s great for us,” Patel said. “Because it allows us to create these machine learning data sets on these novel proteins, but it also gives access to new drug discovery programs in completely new areas of biology.”