The Terray Therapeutics lab is a symphony of miniature automation: Robots whirl around, delivering tiny tubes of liquid to each station. Scientists in blue coats, sterile gloves and safety glasses monitor the machines.
But the real action happens at the nanoscale: proteins in solution bind to chemical molecules held in tiny wells inside custom-made silicon chips the size of tiny muffin tins. Every interaction is recorded millions of times each day, generating 50 terabytes of raw data every day — the equivalent of more than 12,000 movies.
The lab, located in Monrovia, California, and covering an area roughly two-thirds the size of a football field, is a data factory for artificial intelligence-enabled drug discovery and development, part of a wave of new companies and startups using AI to produce more effective medicines, faster.
The two companies are reinventing drug discovery by using new technologies that can learn from and generate answers from vast amounts of data. They're moving the field from painstaking craftsmanship to more automated precision — a change driven by AI that learns and gets smarter.
“AI can work and be very good if you have the right kind of data,” said Jacob Berlin, co-founder and CEO of Telei.
Most of the early business uses of generative AI, which can churn out everything from poetry to computer programs, have been to help alleviate the drudgery of routine office work, customer service and code writing, but drug discovery and development is a huge industry that experts say is ripe for AI transformation.
According to consulting firm McKinsey & Company, AI represents a “once-in-a-century opportunity” for the pharmaceutical industry.
Just as popular chatbots like ChatGPT are trained on text from the internet, and image generators like DALL-E learn from vast amounts of images and videos, AI for drug discovery relies on data — and very specific data like molecular information, protein structures, and measurements of biochemical interactions. Like matching a chemical key to the right protein keyhole, AI learns from patterns in the data to suggest useful drug candidates.
Because AI for drug development is based on precise scientific data, it is much less likely to produce harmful “hallucinations” than more extensively trained chatbots, and potential medicines must undergo extensive testing in laboratories and clinical trials before being approved for patients.
Companies like Terray are building large, high-tech labs that generate information to help train AI, allowing it to experiment rapidly and identify patterns to predict what will work.
Generative AI can digitally design drug molecules. The designs are then converted into physical molecules in a fast-track automated lab and tested for interactions with target proteins. The results (positive or negative) are recorded and fed back into the AI software to improve the next design, accelerating the entire process.
Some AI-developed drugs are in clinical trials, but they are still in the early stages.
“Generative AI is transforming the field, but the drug development process is complex and very human,” says biochemist David Baker, director of the Institute for Protein Design at the University of Washington.
Drug development has traditionally been an expensive, lengthy, hit-or-miss endeavor. Studies vary widely on the costs of designing a drug and taking it through clinical trials to final approval; however, the total cost is estimated to average $1 billion. Development takes 10 to 15 years, and almost 90% of candidate drugs that enter human clinical trials fail, usually due to lack of efficacy or unexpected side effects.
Young AI drug developers are working to use their technology to improve those odds while reducing time and costs.
The most stable source of funding is the big pharma companies, which have long acted as partners and bankers to smaller research ventures. Today's AI drugmakers are typically focused on accelerating the preclinical phase of development, which traditionally takes four to seven years. Some even try to get into clinical trials themselves. But at that stage, big pharma usually takes over and conducts costly human trials that take another seven years.
For established pharmaceutical companies, a partnering strategy is a relatively low-cost means of leveraging innovation.
“For them, it's like taking an Uber to go somewhere instead of buying a car,” said Gerald Ubagus Carrion, a former biotechnology investment banker at Bank of America Securities.
Big pharma pays their research partners for milestone achievements towards potential drug candidates – amounts that can amount to hundreds of millions of dollars over several years – and then royalty revenues flow in if the drug is eventually approved and commercially successful.
Companies like Terray, Recursion Pharmaceuticals, Schrödinger, and Isomorphic Labs are pursuing breakthroughs, but broadly speaking, there are two different paths: those that build large labs and those that don't.
Isomorphic, a drug-discovery company spun out of tech giant Google's core AI group, DeepMind, is betting on the capabilities of its software, embracing the idea that the better the AI, the less data it needs.
In 2021, Google DeepMind released software that accurately predicts how chains of amino acids will fold into proteins. These three-dimensional shapes determine the protein's function. Since proteins drive the behavior of all living organisms, this has boosted biological understanding and even aided in the discovery of new drugs.
Last month, Google DeepMind and Isomorphic announced that their latest AI model, AlphaFold 3, can predict molecule-protein interactions, a further step in drug design.
“We're focused on computational approaches,” says Max Jaderberg, Isomorphic's chief AI officer, “and we think the potential to unlock is enormous.”
Terray, like many other drug development startups, is the byproduct of combining years of scientific research with recent developments in AI.
Berlin, the CEO, has a PhD in chemistry from Caltech and has pursued advances in nanotechnology and chemistry throughout his career. Terray grew out of an academic project that began more than a decade ago at City of Hope Cancer Center near Los Angeles, where Berlin led a research group.
Tellay is focused on developing small molecule drugs — essentially any medicine that can be taken in a pill, like aspirin or a statin, which are easy to take and cheap to produce.
Terray's sophisticated lab is a far cry from the old days of academia, when data was stored in Excel spreadsheets and automation was a distant goal.
“I was a robot,” recalls Kathleen Ellison, Tellay co-founder and senior scientist.
But by the time Terary was founded in 2018, the technology needed to build industrial data labs had advanced rapidly. Terary has relied on advances from outside manufacturers to produce the microscale chips it designs. Its labs are packed with automated equipment, nearly all of which has been customized thanks to advances in 3D printing technology.
From the beginning, the Terray team knew that AI would be important to make sense of the accumulated data, but the potential of generative AI in drug development only became apparent later — but before ChatGPT hit the big time in 2022.
Narve Mardirossian, a senior scientist at Amgen, became Terei's chief technology officer in 2020, in part because of the wealth of data generated in the company's labs. Under Mardirossian's direction, Terei built a data science and AI team to develop AI models that convert chemical data into mathematics and vice versa. The company has released an open-source version.
Terei has collaboration agreements with Bristol-Myers Squibb and Calico LifeSciences, a subsidiary of Google parent Alphabet Inc. that focuses on age-related diseases. Terms of those deals were not disclosed.
Terray will need more than $80 million in venture capital to expand, said Eli Berlin, Dr. Berlin's brother, who left a private equity job to become co-founder and chief financial officer of the startup and said he believed the technology could open the door to a profitable business.
Tellay is developing new drugs for inflammatory diseases such as lupus, psoriasis and rheumatoid arthritis, and Dr. Berlin said the company expects to have a drug in clinical trials by early 2026.
Drug manufacturing innovations from Terray and its peers can speed things up, but there are limits.
“The ultimate test for us, and for the field as a whole, is whether we can look back in 10 years and say that we've significantly improved our clinical success rates and developed better drugs for human health,” Dr. Berlin said.
