Biomatter's Matthew Thompson on the Era of Enzymes on Demand

In an exclusive conversation ahead of their appearance at SynBioBeta 2026 on May 4–7 in San Jose, Matthew Thompson, Biomatter’s VP of Industrial Biotech, laid out the company's vision for enzyme engineering that breaks with how the industry has organized itself for decades. The Lithuania-based company is defining a new category - enzymes on demand - and betting that AI-first design will reshape which projects get funded inside biotech and pharma R&D groups.

Biomatter has been building toward this moment for years. The company pioneered ProteinGAN in 2019, one of the first generative adversarial networks capable of producing functional protein sequences, and has since folded that work into its Intelligent Architecture™ platform, which pairs generative models with physics-based engines to design enzymes from the bottom up. That approach earned Biomatter a €6.5M seed round in 2024 and recognition as Lithuania's AI Company of the Year. In January, Biomatter became the first Baltic and Lithuanian company to join the AstraZeneca BioVentureHub in Gothenburg, deepening its work in therapeutic protein design alongside its industrial biotech programs.

Thompson, who completed his PhD in biocatalysis with Prof. Turner in Manchester, has spent more than a decade in the enzyme industry - most recently as VP of Enzyme Development & Innovation at EnginZyme - traces his interest in synthetic biology to the founding of the Manchester Institute of Biotechnology's SynBioChem centre around 2014.

"As a chemist by training I was drawn to the idea of performing what seemed like magic, producing an endless array of chemicals from sugar," he said. "I was already working with enzymes at the time, but the idea that we could engineer them to work directly on sugar, instead of petrochemical-derived substrates, was very exciting."

That early enthusiasm has hardened into a specific commercial thesis. At SynBioBeta, Thompson plans to walk through the levers any enzyme engineering company has to pull to win, of which de novo design is only one. The harder problem, in his view, is that too many enzyme programs never get greenlit in the first place. Legacy approaches are structurally limited — multi-year timelines, uncertain outcomes, and a ceiling set by what nature already evolved. Faced with that, the most ambitious programs get shelved before they start.

"Enzyme engineering has always been gated by the economics of building and maintaining deep internal capability," Thompson said. "We think that's the wrong model. A specialized platform  running thousands of campaigns compounds learning in a way no single program can, and we pass that advantage directly to the customer as a working enzyme - on demand. Programs that couldn't be justified before become straightforward to greenlight. That's not a threat to internal teams, it's what frees them to focus on the problems only they can solve."

The argument is more than rhetorical. Biomatter's work with Kirin Holdings on human milk oligosaccharide production shows what compressed timelines look like in practice. The team redesigned the substrate binding pocket of α1,2-fucosyltransferase to selectively produce lacto-N-fucopentaose I - the second most abundant fucosylated HMO in breast milk - while suppressing the side product 2′-fucosyllactose. The new enzyme was delivered in under a month, and the strain expressing it accumulated only trace amounts of 2′-FL, a critical requirement for industrial-scale fermentation. It is what enzymes on demand looks like in practice — something a resource-intensive legacy program would have struggled to reach.

Speed alone, Thompson notes, isn't the differentiator - it's what speed unlocks.

"The enzymes we design don't exist in nature, and we deliver them in weeks," Thompson said. "That's not just a faster version of the old model. When the timeline collapses, the economics of entire programs flip. Things that couldn't be justified at multi-year scope become obvious bets when they take weeks, and that changes which products actually get built."

Over the next twelve months, Thompson wants to broaden the kinds of partnerships Biomatter takes on - particularly the ambitious, high-conviction programs where the biology was never the obstacle, only the upfront cost and the timeline. Protein binders were where AI made its first mark in the industry; enzymes are the broader and harder problem, and in his view, the bigger prize. The organizations that pull ahead in the next decade of biomanufacturing will be the ones that pair strong internal teams with a new way of sourcing enzymes — something you access, not something you build.