Artificial intelligence is increasingly impacting the field of chemistry. ReactWise, a company based in Cambridge, U.K. and backed by Y Combinator, is utilizing AI to accelerate chemical manufacturing processes, a critical phase in bringing new pharmaceuticals to the market.
Once a promising compound is identified in the laboratory, pharmaceutical companies need to produce larger quantities to conduct clinical trials. ReactWise proposes to intervene at this stage with its “AI copilot for chemical process optimization.” This AI tool reportedly accelerates the traditional trial-and-error method of determining the optimal drug production process by a factor of 30.
According to co-founder and CEO Alexander Pomberger, drug manufacturing is akin to cooking, requiring the formulation of a high-purity, high-yield recipe. Historically, the industry has relied on trial-and-error or the expertise of staff for process development. By incorporating automation, ReactWise aims to reduce the number of iterations needed, thereby shortening the time to establish a robust manufacturing method.
The startup is ambitiously targeting a future where its AI achieves “one shot prediction,” meaning the ability to predict the ideal experimental setup immediately, without multiple iterative cycles. This breakthrough is anticipated within the next two years, according to Pomberger.
ReactWise’s machine learning models promise substantial savings by minimizing the iterative phase of drug development. The inspiration for this innovation stems from Pomberger’s background as a chemist in Big Pharma, where he noted the tedious nature of the industry’s trial-and-error methods. The company’s offering consolidates five years of academic research into a user-friendly software platform, resulting from Pomberger’s doctoral research on the automation of chemical synthesis using robotic workflows and AI.
ReactWise’s product is underpinned by “thousands” of reactions conducted in its labs to gather the data necessary for its AI predictions. These experiments utilize a high throughput screening method to examine 300 reactions simultaneously, hastening the process of data collection. Pomberger notes the startup focuses on a finite number of chemical reactions commonly used in the pharmaceutical industry, which are essential for training foundational reactivity models that understand chemistry.
Since last August, ReactWise has been documenting reaction types to train its AI, aiming to compile 20,000 chemical data points that encompass the most crucial reactions. Traditional data collection for a single data point takes a chemist one to three days, which is both expensive and challenging.
The company is concentrating on manufacturing methods for small molecule drugs, which can be utilized in treatments for various diseases. Additionally, ReactWise’s technology has the potential to be applied in other fields, as it is also collaborating with companies involved in polymer drug delivery.
ReactWise’s automation strategy includes software that interfaces with robotic lab equipment to enhance precision in drug manufacturing. While the startup does not produce robotic lab equipment, it offers software solutions to operate such devices if available to clients.
Founded in July 2024, ReactWise currently has 12 software pilot trials with pharmaceutical companies and expects some to convert into full-scale deployments later this year. Although specific company names remain undisclosed, the trials include collaborations with major pharmaceutical firms.
ReactWise has revealed the full details of its pre-seed funding round, totaling $3.4 million. This includes previously publicized support from YC ($500,000) and an Innovate U.K. grant of nearly £1.2 million (around $1.6 million). The remaining $1.5 million comes from unnamed venture capitalists and angel investors.
While ReactWise focuses narrowly on a specific portion of the drug development process, Pomberger highlights the considerable impact that speed improvements in this area can have, potentially shortening the overall drug development timeline by an average of 60%. This acceleration could significantly reduce the typical drug development period of 10 to 12 years.
Other startups are also exploring AI applications across different stages of drug development, including the initial chemical discovery phase, potentially leading to compounding effects in automation and innovation. Within the drug manufacturing domain, Pomberger argues that ReactWise is a leader due to its early focus on this area.
ReactWise competes against legacy software that uses statistical methods, such as JMP, and a few emerging AI-driven drug manufacturing firms. However, Pomberger claims that ReactWise’s competitive advantage lies in its high-quality datasets of chemical reactions, which are generated in-house, allowing pretrained models to offer immediate process recommendations. This feature is unique in the industry, as most competitors provide software that relies on user inputs, whereas ReactWise offers solutions based on extensive pre-existing laboratory work.
