Artificial intelligence keeps things chemically agitating. WIT: Y Combinator Backed Cambridge, UK-based ReactWise uses AI to speed up chemical manufacturing. This is an important step in bringing new drugs to the market.
Once promising drugs are identified in the lab, pharmaceutical companies need to be able to produce far more materials to conduct clinical trials. This is where ReactWise offers to intervene in “AI co-pilot for optimizing chemical processes.” It says it will accelerate the standard trial-and-error based process by 30 times that you know the best way to make a drug.
“Medicine is like cooking,” co-founder and CEO Alexander Pomberger (pictured on the left, with co-founder and CTO Daniel Wigg) said in a call with TechCrunch. “You need to find the best recipe for making medicines with high purity and high yield.”
For years, the industry has relied on trial and error in this “process development” or summarizing it in staff expertise, he said. Adding automation to the mix is a way to reduce the number of repeated cycles needed to land on a solid recipe for manufacturing drugs.
Startups believe that AI can provide “one-shot predictions” that allow AI to “predict ideal experiments” almost immediately. Data from each experiment do not require multiple iterations supplied to further refine predictions, in the near future (2 years from now, Pomberger's bet).
Startup machine learning AI models can still provide massive savings by reducing the amount needed to overcome this bit of the drug development chain.
Cutting boredom
“The inspiration for this is that I was a chemist in training and worked at Big Pharma. And I saw how boring and trial and error the whole industry is,” he said.
Supporting Reactwise's products are “thousands of” reactions that startups have performed in their labs to capture data points to deliver AI-driven predictions. Pomberger said the startup used the “high-throughput screening” method in its lab, allowing 300 responses to be screened at a time, speeding up the process of capturing all this training data for AI.
“In Pharma, there are reactions that are used again and again, one or two reaction types,” he said. “What we're doing is that we have labs that generate thousands of data points for these most relevant reactions and train basic reaction models on our side, and those models can fundamentally understand chemistry. And if the client's pharma company needs to develop a scalable process, then we don't have to start from scratch.”
The startup began this process of capturing reaction types to train AIS last August, saying that Pomburger will be completed by summer. We work across 20,000 chemical data points to “cover the most important reactions.”
“It usually takes a chemist, a day to get one data point using the traditional method,” he said. Getting a single data point is extremely difficult. ”
So far, it has focused on the manufacturing process of “small molecule drugs,” and Pomberger says it can be used in medicines targeting all types of disease. However, he suggested that the technology could also be applied to other areas, noting that the company is working with two material manufacturers in the development of polymer drug delivery.
Reactwise's automation play also includes software that allows you to interface with robot lab equipment and dial further precise manufacturing of drugs. However, to be clear, we focus purely on the selling of software. It is not the manufacturer of the robot lab kit itself. Rather, they have added another string to the bow so that if a customer gets such a kit, they can provide to drive robot lab equipment.
Founded in July 2024, the UK startup has 12 pilot trials of software running by pharma companies. Pomberger said the first conversion (hoping to convert to a full-scale deployment of subscription software later this year.” And while it hasn't revealed the names of all the companies it is working with yet, Reactwise said these trials will include several major pharma players.
Pre-seed funds
Reactwise discloses details of pre-seed pay raises. This totals $3.4 million.
The figures include previously disclosed support from YC ($500,000) and nearly £1.2 million in UK grants (approximately $1.6 million). The remaining funds (approximately $1.5 million) come from unknown venture capitalists and angel investors, and Reactwise says it is “committed to moving forward with AI-driven sustainable drug manufacturing.”
While Reactwise focuses fairly narrowly on certain parts of the drug development chain, Pomberger said acceleration here could make a meaningful difference in reducing the time it takes for patients to acquire new medicines.
“Let's take a look at the typical period of drugs from start to launch: 10-12 years. Process development takes 1.5-2 years. And if you can basically speed up your workflow here, you can see how effective it is if you can reduce your average by 60%,” he pointed out.
At the same time, other startups are applying AI to various aspects of drug development, including identifying interesting chemicals in the first place, which could potentially have more complex effects as more automation innovations fold.
But when it comes to drug manufacturing, specifically, Pomberger claims that Reactwise is in front of the pack. “We were actually the first to work on this,” he said.
Startups compete with legacy software using statistical approaches such as JMP. He also said there are several other applications of AI to speed up drug manufacturing, but ReactWise's access to high-quality datasets on chemical reactions is competitive.
“We have the capabilities of these high quality datasets and are the only ones that are currently generating,” he said. “Most of our competitors offer software. Our clients are basically asked to give instructions based on input.
“But from our side, we offer these pre-secured models. They are very powerful because they understand chemistry at its core. And the idea is what we really say to our clients: “This is my reaction of interest, the start of a hit, and we have already provided recommendations for the process from the first day, based on all the pre-work we did in the lab. And that's something no one else has done.”