A team of researchers at the University of Cambridge have designed a new approach to the preclinical drug development process using machine learning, which has the potential to speed up drug development and move ideal drug candidates further down the drug development pipeline. The team’s work is published in two separate articles in Nature Chemistry and Nature Communications.
Developing new drugs is a complex, expensive process, and it starts with a process of identifying molecules and making predictions about how they will react. The right molecules can move forward in the development process and potentially lead to the development of actual pharmaceuticals. However, this is a trial and error process, making it hard to accurately predict the best molecules to move forward in the process.
Enter machine learning, a tool that is being leveraged with increased frequency in medicine and clinical development work. Specifically, the team leveraged a machine learning approach that’s similar to genomics: experiments done to predict how molecules will react are combined with a machines ability to process data and learn to recognize patterns in how certain chemicals react. The machine can, for example, identify important correlations between different agents in a reaction and provide this data quickly to researchers. To date, they’ve compiled roughly 39,000 different molecular reactions with implications for clinical development. The team has dubbed their approach the chemical reactome.
To build on this approach, the University of Cambridge team has gone a step further with machine learning and designed a tool that can help these molecules further along in the drug development process. Specifically, the machine approach allows researchers to manipulate molecules by introducing changes to certain areas of the molecule. The ability to do this allows researchers to modify molecules without having to build a molecule from scratch in the lab. This approach, the team hopes, will only improve the speed at which drugs are developed and introduced to the clinical pipeline.
Sources: Science Daily; Nature Chemistry; Nature Communications
