Article image

AI machine learning could revolutionize new drug discovery

The advent of AI machine learning in the realm of pharmaceutical drug synthesis marks a major shift in drug discovery and development. 

A team of researchers at the University of Cambridge, in collaboration with Pfizer, have developed a transformative approach to drug research that merges automated experiments with artificial intelligence (AI).

This method could significantly expedite the creation of new drugs.

AI is changing the game for drug companies

Historically, the process of predicting chemical reactions has been largely based on trial-and-error, leading to frequent failures. 

Traditional techniques involve simulating electrons and atoms in simplified models, which are not only computationally demanding but also prone to inaccuracies. 

The Cambridge team, inspired by genomics, has taken an AI data-driven route to drug synthesis and understanding chemical reactivity. Their new approach, which they have named the chemical “reactome,” is a game-changer.

“The reactome could change the way we think about organic chemistry,” said study lead author Dr. Emma King-Smith from Cambridge’s Cavendish Laboratory. 

“A deeper understanding of the chemistry could enable us to make pharmaceuticals and so many other useful products much faster. But more fundamentally, the understanding we hope to generate will be beneficial to anyone who works with molecules.”

Chemical reactome

The chemical reactome identifies correlations between reactants, reagents, and the reaction’s performance from a dataset of over 39,000 pharmaceutically relevant reactions.

It also highlights gaps in existing data. The method leverages high-throughput automated experiments to generate this data.

“High throughput chemistry has been a game-changer, but we believed there was a way to uncover a deeper understanding of chemical reactions than what can be observed from the initial results of a high throughput experiment,” explained Dr. King-Smith.

“Our approach uncovers the hidden relationships between reaction components and outcomes,” said Dr Alpha Lee, who led the research.

“The dataset we trained the model on is massive – it will help bring the process of chemical discovery from trial-and-error to the age of big data.”

Efficient drug design using AI

A related study, published in Nature Communications, showcases the team’s development of an AI machine learning approach for precise molecular transformations in drug synthesis. This method enables chemists to make specific changes to the core of a molecule. 

The technique is comparable to a last-minute design tweak, without starting from scratch. Such flexibility is crucial for efficient drug design, particularly for late-stage functionalization reactions, which are often unpredictable and challenging to control.

“Late-stage functionalization can yield unpredictable results and current methods of modeling, including our own expert intuition, isn’t perfect. A more predictive model would give us the opportunity for better screening,” said Dr. King-Smith.

The AI machine learning model, pre-trained on extensive spectroscopic data, is capable of predicting reaction sites and how these vary under different conditions.

This approach overcomes the issue of scarce data in late-stage functionalization, enabling accurate prediction of reactivity sites on diverse drug-like molecules.

Predicting intricate transformations 

“We pretrained the model on a large body of spectroscopic data – effectively teaching the model general chemistry – before fine-tuning it to predict these intricate transformations,” said Dr. King-Smith. 

The team experimentally validated the model on a diverse set of drug-like molecules and was able to accurately predict the sites of reactivity under different conditions.

“The application of machine learning to chemistry is often throttled by the problem that the amount of data is small compared to the vastness of chemical space,” said Dr. Lee.

“Our approach – designing models that learn from large datasets that are similar but not the same as the problem we are trying to solve – resolve this fundamental low-data challenge and could unlock advances beyond late stage functionalization.”  

The study is published in the journal Nature Communications

Like what you read? Subscribe to our newsletter for engaging articles, exclusive content, and the latest updates.


Check us out on EarthSnap, a free app brought to you by Eric Ralls and

News coming your way
The biggest news about our planet delivered to you each day