In a laboratory near downtown Seattle on the shores of Lake Union, Hetu Kamisetty is helping build what he says is the biotech company of the future.
If talent, money and novel technology are any indication, he and his colleagues have a good shot at it.
Kamisetty heads the Seattle location of Xaira Therapeutics, an AI-focused startup launched in April with more $1 billion from investors.
The lab is populated by researchers skimmed from the nearby Institute for Protein Design, the lauded University of Washington research hub and spinout machine.
The founding members of Xaira’s molecule design and AI team, which Kamisetty leads, are all IPD alums.
Nathaniel Bennett helped develop RFDiffusion, an AI-powered tool from the IPD used to design proteins with drug-like properties. Justas Dauparas, a soft-spoken soccer-playing Lithuanian, is one of the scientists behind another IPD model, ProteinMPNN. Buwei Huang is a recent IPD graduate student and Philip Leung is a former postdoc.
Xaira was founded with the aim of building on IPD models like RFDiffusion and ProteinMPNN, and going beyond them to develop new therapeutics based on proteins and other molecules. Xaira’s scientists integrate molecular design with other AI-based approaches to model biology and drug development across the drug development pipeline.
“If you think of drug discovery, it connects the world of chemistry and molecules to the world of disease and patients. You need to span that entire space in order to make meaningful progress,” said Kamisetty, a Xaira co-founder.
The market is potentially vast. Biologics like protein-based therapeutics accounted for a third of drug approvals in 2022.
Riding the AI wave
Kamisetty was most recently a scientist for Facebook’s generative AI team. The move to Xaira brought him full circle. Ten years ago he was a senior fellow in the lab of David Baker, now the head of the IPD, back when protein design was a nascent endeavor.
Now, the field is booming, and the IPD is at its center. The IPD has spun out five companies in the last five years, and racked up successful exits for its prior launches. PvP Biologics sold for $330 million in 2020 and vaccine design company Icosavax was bought for $1.1 billion in February. Baker has co-founded 21 companies, including Xaira.
Scientific accolades have come in tandem. In 2021, Baker took home the prestigious Breakthrough Prize in Life Sciences, and he and his colleagues shared Science magazine’s Breakthrough of the Year award with DeepMind.
Science recognized the researchers for their deep learning tools to predict how proteins fold into three-dimensional shapes. The feat stunned scientists and set the stage for later “generative” models like ProteinMPNN, RFDiffusion and, most recently, RFantibody.
These generative models dream up new proteins based on input about how they should look or act, or what other molecules they should interact with. The process is roughly similar to how ChatGPT or DALL-E generate text or images based on prompts.
The models help researchers create potential therapeutics, such as proteins that bind to and neutralize cellular molecules involved in disease.
Designs beyond nature’s
Dauparas joined the IPD with fresh graduate degrees in mathematics and biophysics as interest was building in a now-famous 2017 publication, “Attention is all you need.” The publication later led to today’s popular generative AI models.
“When that came out, we started to think, ‘this could be applied to proteins,’” he said. The field now offers the possibility of building therapeutic proteins that surpass nature’s designs.
Traditionally, new protein-based therapeutics originate with nature templates, such as antibodies developed in mice. Nature’s antibodies can be forged into drugs that zero in on cellular targets. But such antibodies don’t always work. Some targets are intransigent and fail to yield new therapeutics — in industry parlance, these targets are “undruggable.”
Molecular design has the potential to successfully hit these difficult targets, said Xaira researchers. The IPD’s technology has already been shown to yield new designs not found in nature. These include entirely new antibodies and miniproteins that bind to human biomarkers and other biological molecules.
Xaira is mum about its therapeutic areas of focus. But the company has a big goal that makes it stand out, said Huang: “Conquering undruggable targets.”
That vision has helped to attract some big-name venture capitalists. Co-founders include Robert Nelsen, managing director of Arch Venture Partners, known for making big bets in biotech, and Vikram Bajaj, managing director of Foresite Capital.
Xaira was jointly incubated by Arch and Foresite Labs, and funded by a host of other investors. Former Stanford president and Genentech scientist Marc Tessier-Lavigne was tapped as CEO.
From desktop to benchtop
Most of Xaira’s 80 employees work from its headquarters in the Bay Area, with a handful in London and 15 people in Seattle.
The IPD’s UW building is visible across the lake from Xaira’s space in the Dexter Yard life sciences building.
“There’s very few places that marry tech and biotech in the way that Seattle is set up to do,” said Kamisetty, who aims to build the Seattle team to about two dozen employees by the end of the year. “For protein design, this is the place to be.”
On a recent visit to the Xaira lab, the Seattle team is decked out in company-branded T-shirts from a photo-op earlier in the day. The colorful loops and helices of protein designs light up monitors where scientists dream up new molecules. Steps away, laboratory researchers synthesize and test the designs using a phalanx of high-throughput machines.
Laboratory tests assess how well the proteins stick to specific cellular targets, and probe other properties such as stability. The data are quickly fed back into the protein models, enabling the next iteration of molecular design.
“We want to accelerate the feedback loop from generating experimental data, back to the model,” said Huang. “The tight connection is very important.”
After researchers design molecules and run initial tests, promising candidates head to the Bay Area labs. There, they are subject to further testing and refinement, with the aim of readying them for clinical trials.
“From molecules, to biology and patients, we are building models and collecting data,” said Kamisetty. “We think of the data, models, and iteration across the entire spectrum.”
Xaira was inspired by the IPD’s close connection between its computational and laboratory teams, said Kamisetty. That’s been key to the IPD’s success, he said, along with quick adaptation to emerging AI methods.
A race to the top
Xaira’s neighbors in Dexter Yard include cell therapy-focused protein design startup Outpace Bio and IPD spinout Monod Bio, which recently released its first product for molecular assays.
Dauparus, Huang and other workers in the building get together for soccer, and pickup basketball in the building’s basement court.
Outside Dexter Yard, Xaira is jostling with a host of competitors. These include AI-powered startup Insitro, which raised $400 million in 2021, and Generate:Biomedicines, which is fueled by close to $750 million and is using its own protein design model, Chroma. In August, Exscientia merged into Recursion, yielded a machine-learning focused company with $850 million in cash on hand.
Meanwhile, big pharma companies from Eli Lilly and Company to Bristol Myers Squibb are beefing up their AI capabilities.
The biotech landscape is shifting, said Kamisetty. The big biotech companies of today will not be the big companies of tomorrow. The winners, he said, will be “digitally native.”
“To be a successful biotech company you need to build on AI,” said Kamisetty. “My take is that five or ten years from now, this will just be the way biotech and drug discovery is done.”
https://ift.tt/ELymhUH August 14, 2024 at 03:14PM GeekWire
Post a Comment
0Comments