fb-pixelWill artificial intelligence help build better protein drugs? This scientist quit his job to prove it can. - The Boston Globe Skip to main content

Will artificial intelligence help build better protein drugs? This scientist quit his job to prove it can.

Chris Bahl shut down his academic lab to focus on a biotech startup, AI Proteins, which has raised $18.2 million.

Chris Bahl, founder, president, and chief scientific officer of Andover-based AI Proteins, is using artificial intelligence to develop new protein therapies from scratch.David L. Ryan/Globe Staff

Chris Bahl is a self-proclaimed protein geek. He’s spent the better part of 10 years studying the intricate structures of proteins — complex molecular machines responsible for nearly all facets of life, from metabolism to movement. His goal was to create completely synthetic proteins using computer programs, and to open up a new avenue for making medicines.

His work, which relies heavily on recent advances in artificial intelligence and synthetic biology, had taken him from one prestigious protein lab in Seattle to another one in Boston. “Things were going super well,” he said.

But last fall, after a stream of queries from drug companies hoping to partner with him, Bahl decided to cancel his six research grants, close shop at his four-year-old academic lab, and launch a biotech startup to focus on making protein therapies full time. “It was time to fly,” he said.

Advertisement



After a year of working in stealth mode, Bahl’s startup, AI Proteins, emerged on Thursday with $18.2 million in seed funding led by venture capital firms Cobro Ventures and Lightchain Capital. The money will help the startup, where Bahl is president and chief scientific officer, refine its protein creation technology and further progress on a dozen experimental therapies it’s already developing for immune diseases and cancer.

Lightchain, a St. Louis firm that invests in life science and software companies, was shocked by how quickly the startup could design and optimize the potential drugs. “It’s like nothing we’ve seen before. It’s almost unbelievable,” said Drew Dennison, managing director of Lightchain, who is serving as chief executive of AI Proteins. “It feels like there’s no limit to the technology.”

The startup is the latest entrant to a large and growing field of biotech firms making bold and often untested claims about how artificial intelligence and machine learning will help them make better drugs more quickly and at lower cost.

Advertisement



“There’s no question that this space is incredibly overhyped,” said Peter Sorger, a professor of systems biology at Harvard Medical School. “Investors want to imagine that drug discovery is going to be easy now, and that will never be true.” But if artificial intelligence can help make drugs even 10 percent better or faster, “that is really meaningful,” he added.

A view on the computer of a miniprotein molecular model at the Andover-based biotech company AI Proteins.David L. Ryan/Globe Staff

Most biotechs focused on artificial intelligence are using it to make small molecule compounds, typically administered as pills and traditionally designed by chemists. Other companies are using machine learning to optimize existing proteins. But AI Proteins is one of the few trailblazing a technique called “de novo protein design” that promises to break free from the constraints of what nature has already made.

“This is creating proteins totally from scratch with no evolutionary history to anything that’s existed before,” Bahl said. “It’s the field of synthetic biology finally becoming fully synthetic.”

Bahl said his startup writes its own software for designing proteins on supercomputers. It makes those proteins in the lab, runs tests on them with help from robots, and then feeds the data from those experiments back into its computers for the next round of designing, building, and testing.

Proteins therapies, commonly prescribed for cancer, immune diseases, and rare genetic conditions, are part of an enormous global market whose value approached $284 billion in 2020 and is predicted to nearly double to $567 billion by 2030, according to a report from QY Research Medical.

Advertisement



Many of these proteins have already been tweaked by scientists to turn them into better drugs. But some scientists say that de novo protein design will lead to medicines that can’t be made from optimizing existing proteins.

“We’re not starting from something good and making it better. We’re creating perfection, ideally, from the start,” said David Longo, chief executive of the New York biotech startup Ordaōs, which raised $5 million in seed funding in August to design new proteins with machine learning.

Bahl cut his teeth making proteins when he joined David Baker’s lab at the Institute for Protein Design in Seattle in 2012. Bahl led the development of techniques for making a class of small synthetic proteins dubbed miniproteins capable of binding targets implicated in cancer or other diseases.

Senior research associate Isabelle Moczygemba working at the Andover-based biotech company AI Proteins.David L. Ryan/Globe Staff

Buoyed by the high-profile miniprotein research published in 2016, Bahl headed east the following year to start a lab of his own at Timothy Springer’s newly formed Institute for Protein Innovation in Boston where he combined software, robotics, and synthetic biology to develop more miniproteins.

Unlike most proteins, which must be infused or injected, Bahl believes that the compact and sturdy structures of miniproteins could protect them from digestive juices and allow people to take them as a pill. Miniproteins could also be easier to manufacture and store than traditional proteins, and reach deeper into the body.

“They’re small and they go everywhere in the body really rapidly, so we can start to tackle diseases that have historically been very challenging for antibodies,” Bahl said.

Advertisement



As the approach started attracting more attention from drug companies, Bahl began talking to investors in July 2021 about forming his own miniprotein company. Just a few months later, Bahl closed his academic lab and launched AI Proteins in Andover — he couldn’t find lab space on short notice in Boston.

Ten people from Bahl’s academic lab joined his company, and the headcount at AI Proteins has since doubled. Later this month, the startup is moving back to Boston.

In theory, the company could make miniproteins for just about any disease. “One of the big challenges we’ve struggled with over the past year is trying to narrow it down,” Bahl said. To start, the company is making therapeutics for inflammatory diseases that hit their targets more precisely to reduce side effects common in commercial drugs.

AI Proteins already has a handful of competitors making miniproteins, including Ordaōs. The Somerville biotech startup Generate Biomedicines, which uses machine learning for protein engineering, also has ambitions for de novo protein design, including miniproteins.

Research being conducted at biotech company AI Proteins, an Andover-based startup where artificial intelligence is being used to develop new protein therapies from scratch. David L. Ryan/Globe Staff

Generate has raised at least $470 million from investors and partners since it was launched in 2018 by Flagship Pioneering, the life science investment firm behind Moderna. The startup’s lead programs are antibody drugs for COVID-19 and asthma, which chief executive Michael Nally hopes will be more effective and more convenient to take than existing drugs.

Improving on existing drugs allows the firm to minimize the risks associated with new medicines. “That’s probably the lower hanging fruit compared to designing something completely from scratch,” said Nicholas Polizzi, a protein designer at Dana Farber Cancer Institute. “But Generate is the company that any new startup in this field has to set themselves apart from.”

Advertisement



Developing new proteins is not without risk. If the proteins truly look completely different from anything in nature, there’s a strong chance we may develop immune reactions to them — which could foil the drug’s effectiveness and potentially cause dangerous side effects. “That’s a very important downside that needs to be overcome,” Springer said.

Bahl wouldn’t disclose the specific diseases or targets of its initial drug programs, but he did say the company is preparing to test its three lead candidates in animals. Clinical trials could come a year or two after that.

“As soon as there’s a proof of concept that these are safe, effective medicines that can do stuff that hasn’t been done before, I think lots of people are going to be running to catch up and move into this space,” Bahl said.


Ryan Cross can be reached at ryan.cross@globe.com. Follow him on Twitter @RLCscienceboss.