A Failed Control Experiment Led Cambridge Chemists to a New Way of Building Drugs

University of Cambridge

David Vahey was running a control experiment. The PhD researcher at St John's College, Cambridge, had been testing a photocatalyst - a light-activated molecule meant to drive a chemical reaction. As part of standard scientific practice, he removed the catalyst to confirm it was necessary. The reaction worked anyway. In some cases, it worked better.

At first, the unusual product looked like a mistake. Instead of discarding it, Vahey stopped and investigated. That decision, published March 12 in Nature Synthesis, has yielded a new type of chemical reaction that could substantially change how drug molecules are built and modified.

The problem with early-stage chemistry

One of the most fundamental steps in building drug molecules is forming carbon-carbon bonds - the links that underpin virtually all organic chemistry. A classic method for doing this is the Friedel-Crafts reaction, which uses strong acids or metal catalysts under harsh conditions. The harshness is the problem: these reactions can only happen early in the drug manufacturing process, when the molecule is still simple enough to withstand the rough treatment. After that, chemists must perform many additional steps to build up the final drug.

This means that if a pharmaceutical scientist wants to test a small modification to an existing drug - swapping one chemical group for another to see if it improves efficacy or reduces side effects - they often have to start from scratch. Rebuilding large parts of a complex molecule just to change one small feature can take months.

The anti-Friedel-Crafts approach

The Cambridge team's new reaction inverts this pattern. It forms carbon-carbon bonds under mild conditions, powered by an LED lamp at ambient temperature, through a self-sustaining chain process. No toxic metal catalysts. No extreme temperatures or pressures. No expensive reagents.

Because the conditions are gentle, the reaction can be applied to complex, nearly finished drug molecules without destroying the delicate structures that took months to build. Chemists call this late-stage modification - and it is one of the most sought-after capabilities in modern medicinal chemistry.

The reaction is also highly selective, meaning it can alter one part of a molecule while leaving other sensitive regions untouched. In practical terms, this lets pharmaceutical scientists make targeted changes at the end of the development process rather than rebuilding from the beginning.

Serendipity, then science

The discovery followed a pattern familiar in the history of chemistry. Vahey's supervisor, Professor Erwin Reisner, described the moment: failure after failure, then finding something unexpected in the mess. Reisner, whose group at Cambridge is known for developing photosynthesis-inspired systems, said recognizing the value in the unexpected is probably one of the key characteristics of a successful scientist.

Once the team understood the underlying chemistry, they collaborated with researchers at Trinity College Dublin to build machine-learning models that could predict where the reaction would occur on molecules that had never been tested in the lab. By learning patterns from established chemistry, the AI could simulate reactions before they were run, helping identify the most promising candidates faster.

The team also partnered with AstraZeneca to test whether the method could meet the practical and environmental demands of large-scale pharmaceutical development. They demonstrated the reaction across a wide range of drug-like molecules and showed it could be adapted to continuous-flow systems increasingly used in industry.

Environmental and efficiency gains

The practical benefits extend beyond speed. By eliminating metal catalysts and harsh chemicals, the reaction reduces toxic waste. Fewer synthetic steps mean less energy consumption and a smaller environmental footprint. This matters in pharmaceutical development, where the industry is under increasing pressure to reduce its environmental impact.

Reisner framed the broader context: transitioning the chemical industry to sustainability is arguably one of the most difficult parts of the whole energy transition. A reaction that builds fundamental chemical bonds without toxic reagents or heavy metals represents a small but meaningful step in that direction.

From one reaction to a toolbox

The study demonstrates one specific reaction type applied to carbon-carbon bond formation. It does not address all the chemical transformations needed in drug development, and it will not replace the full synthetic toolkit that pharmaceutical chemists rely on. Its value lies in adding a new capability - late-stage C-C bond formation under mild conditions - that has been difficult to achieve by other means.

Whether the reaction proves as widely useful as the initial results suggest will depend on how well it performs across the enormous diversity of molecular structures encountered in real drug development programs. The AstraZeneca collaboration is promising but represents early-stage testing, not validated industrial deployment.

The machine-learning component, while useful for prediction, is trained on existing chemical knowledge and cannot predict genuinely novel chemistry. It accelerates exploration within known chemical space rather than expanding the boundaries of what is chemically possible.

And the serendipitous origin of the discovery, while a good story, does not guarantee broad applicability. Many accidental discoveries remain narrow in scope. The true test will come as other research groups attempt to reproduce and extend the work across different molecular contexts.

For Vahey, the significance is practical: giving researchers a new tool in the drug discovery toolbox. What industry and other researchers do with it next - that is where the future impact lies.