AI is a key ingredient of modern formulation science

Formulation science is a complex business. The design and development of complex products such as advanced coatings and cosmetics generates massive data sets that the standard digital toolkit — from electronic lab notebooks to the trusty spreadsheet — struggle to manage. Data can become siloed, limiting the impact of the intelligence it contains.

As R&D teams work to address these data challenges, AI is finding a growing place in product design workflows. And yet, despite AI’s potential, it’s easy to lose track of where these tools will make a real difference. The answer lies not just in the data you produce, but how you use it.

“Traditional statistics is about understanding your design space,” says Tom Whitehead, head of machine learning at R&D AI specialist Intellegens. “But in an industrial setting, normally the question is how to hit performance goals as quickly as possible. And that’s where AI can come in.” Smart use of AI, Whitehead says, can cut the number of experiments required by 50–80%.

“No-one knows the boundaries of what we can do at this point,” says Mark Goulding, programme director for R&D digitization in the Electronics business of Merck KGaA Darmstadt, Germany. “AI-assisted drug design and materials application test processes are fairly well established, but in other fields there’s a lot more human judgement needed. Where AI can come through is interpretation of data, controlling conclusions and outcomes, and being a guide — allowing you to analyse bigger data sets and look for deeper patterns.”

AI’s aptitude for large-scale data management and analysis makes it a natural companion for formulation scientists, who operate in a multifaceted data environment and have traditionally struggled to break down barriers between project teams.

Opportunities for integration

Such an approach doesn’t work in isolation. It requires a deep understanding of science and partnerships to integrate the techniques into all aspects of the business. This is why Intellegens has partnered with Revvity Signals, which has integrated AI into its Signals One platform to improve data management, design of experiments and analytics for its users.

Merck adopted Signals well before the AI boom, when it realized it needed better-quality, structured, traceable and fairer data. After considering the alternatives, including building something in-house, it settled on Signals Notebook because of its cloud working system, and its integration with critical tools such as ChemDraw. “It made things accessible from all our R&D sites around the world,” Goulding says. “So we had 24/7 working in the US, Europe and Asia, making collaboration much easier.”

With the introduction of AI, Goulding can see even greater opportunities with the platform. He offers an example: “You might do a series of experiments with five instrument measurements. There are quite a lot of variables: measurements, samples, test conditions. AI can look at that data and find trends. You want the digitization to take away the manual, repetitive work, allowing scientists more time for quality human input and creative tasks.”

And, when combined with a platform that facilitates cloud collaboration, this streamlining brings entirely new possibilities. “Signals One takes the best tools from both the AI and data management arenas,” Whitehead says, “to give users the best of both worlds.”