Damian Nowak – Adam Bachorz – Professor Marcin Hoffmann | Using Machine Learning to Discover New Medicines

This means that while this immense ocean of candidates is almost certainly hiding an abundance of life-saving medicines, the limited scope of existing search techniques makes them virtually impossible to identify.

Through their research, Damian Nowak and Marcin Hoffman at Adam Mickiewicz University in Poznan, together with Adam Bachorz at the Polish Academy of Sciences, show how the pace of drug discovery could be vastly sped up with the help of artificial intelligence.

Designing new drugs involves identifying molecular groups that can effectively bind with specific target proteins. This can trigger a wide array of biochemical processes, such as immobilising harmful pathogens, or stimulating the immune system to detect and destroy cancer cells.

In their study, Nowak, Bachorz and Hoffmann explored how the discovery of new molecular groups could be aided by machine learning: a branch of artificial intelligence that involves building models designed for specific tasks. These machine-learning models improve their performance through their experience of real data, resulting in an artificial intelligence capable of proposing chemical substances likely to act as powerful drugs.

The team trained their machine-learning model with 121,000 known molecules, allowing it to identify links between molecular structure and drug effectiveness. Once trained, the model identified entirely new groups of molecules, specifically tailored for binding with target proteins that had been specifically selected by the team.

By developing their model further, Nowak, Bachorz and Hoffmann now hope that researchers could ultimately bypass the need to search through vast numbers of candidate drugs individually. If achieved, their new approach could rapidly speed up the pace of drug discovery, ultimately leading to new treatments with the potential to save millions of lives globally.