Since its inception, the chemical industry has fixated on what molecule to make and then on how to make it. Chemify is shaking up that process, enabling scientists to advance more quickly using a novel digital approach.
“If you design a molecule (the traditional way), you don’t know if that molecule can really be made…or made quickly,” says Lee Cronin, PhD, founder and CEO of Chemify. His company’s three-pronged approach, dubbed “chemputation,” aims to be a more dependable, reproducible, and efficient means to transition molecules from design to a physical entity and to scale easily from laboratory to clinical trial or commercial quantities.
The idea blends digitization and robotics with molecule discovery. Chemify combines a new, purpose-made programming language with robotics and artificial intelligence to create what Cronin expects will become “the world’s largest repository of molecules that can be made.”
Chemify converts molecules and the chemical procedures into digital code. Robots can then use that code to manufacture existing and yet-to-be-designed molecules on demand. Quantities can be scaled up by adding bioreactors rather than enlarging reactors.
This technology should result in the rapid creation and production of superior small molecules. As Cronin explains, “We do all the chemistry first, so we only choose molecules we know we can make.” By training the robots to perform the chemistry, this
approach can make two orders of magnitude more reactions than traditional automation (1,000 to 5,000 reactions, Cronin says, compared to the 10-15 reactions that are feasible with existing robots).
In contrast, “Typically, everybody doing in silico drug discovery invents the molecule before they know if they can make it.” The challenge with such an approach, Cronin says, is that drug designers can only go so deep. “Small molecule drug discovery is slowing…some might say failing…because we’re at the bottom (using a diving analogy) and we’re all searching the same space.” By creating a large library of digital molecules, similar to the computational libraries created for biologics, chemists may be able create entirely new, purpose-built chemical entities.
One of the first challenges, Cronin says, “is to convince chemists we’re not taking their jobs. Instead, we’re just giving them a (metaphorical) exoskeleton to let them go deeper into a chemical space.” The difference, he says, is like free diving versus scuba diving. “With scuba, you can go deeper. Likewise, with chemputation, scientists can make more complex molecules.”
There’s also a misperception that Chemify is a manufacturer. Actually, the company is more of a boutique designer, he says. “We’re making complicated molecules for people quickly, that they couldn’t get anywhere else.” The goal complements their customers’ existing capabilities.
Manufacturing facility
Chemify’s first manufacturing facility is scheduled to go online in 2025 in Glasgow. Once that pilot plant is operating smoothly and optimized, Cronin says he may look to build a similar facility in the U.S.
Walking through the planned facility, Cronin says “It won’t look that different from a normal laboratory because we’re using a few hoods, like in a chemistry lab, for air management. In the future, a chemputation facility will look like a server farm with racks of servers and computers.” Here, however, rather than racks of servers, there will be racks of reactors with ducting and fire suppression. “Solvents and solids will move in, and molecules will move out.”
The sweet spot in terms of production is the milligram to gram scale, “but we automate the gram scale. No one else does that,” Cronin maintains.
Using a chemputation approach, scaling up production will be a matter of adding more reactors rather than adding larger reactors. That makes it most cost-effective and easier to scale from lab and preclinical quantities to clinical and commercial quantities, he says.
There also may be a regulatory advantage because standardized robotic units are used. Product output will be comparable from each robot, which avoids the uncertainties associated with up- or down-sizing reactors. “The robot history, code, and chemicals will all be known,” he says, and the log and audit trail are accessible.
The standardized programming language that Cronin developed is an important element in the chemputation approach, he points out. “That programming language works to standardize robots.” Coupled with traceability capabilities, “it enhances performance and reliability, which makes regulatory oversight easier,” he adds.
In-house, for now
While Chemify mines chemical space, the entire chemputation process, including its robotics, will remain in-house. “At the rate at which we’re iterating the technology, it makes no sense (to do otherwise),” Cronin says. Instead, “We want to give partners the ability to co-design with us, making molecules and new drugs at scale.”
Once the chemputation concept is fully mature and outside entities are contributing to the digital molecule knowledge base, the company could become more of a digital developer, Cronin speculates. However, “right now, it’s important for us to make sure the robots work correctly, safely, and traceably.”
This positions Chemify as “an enabler of the design. I view us as the architect and the structural engineer” designing and building actual molecules, he adds.
Disruptive innovation
Cronin formed Chemify in 2022 after developing a chemical description language, cDL (pronounced ChiDL). “Everyone said the language was boring,” he recalls. But, “boring” doesn’t mean “bad.” This purpose-built language, he believed, was an important tool “to make chemistry accessible and reusable,” and thus power the discovery of new drugs.
This vision of digital chemistry resonated with others, prompting Cronin to spin-out the company from the Digital Chemistry Laboratory at the University of Glasgow. “The first employee came on board the first of March, 2022. Now we’re at 120 people, two-and-a-half years on.
“We have a hybrid set of chemical informaticians, chemists, operations engineering, and software developers, all working under one roof,” he says. They’re united around the goal of digitizing chemistry–“this thing that’s never been done before,” he continues.
The potential discoveries in the coming decades, he predicts, “are mind-blowing, and I think that’s got a lot of people excited because they want to contribute to something that has a bigger meaning than just making money.”
In the not-too-distant future, Cronin envisions pharmaceutical companies buying or accessing discovery computer engines to mine chemical space. They could then run their results through Chemify’s chemputation platform to produce physical molecules. “It’s a bit like using a Nvidia chip for openAI,” he explains, in that “Nvidia enables the compute system as well as the model building tools.
“One of the things we’re doing is making sure our customers understand that their intellectual property remains their intellectual property. The only thing Chemify learns (from its clients) is how to do the chemistry. The descriptions of their targets, proteins, and specifications remain completely owned by the customer.”
