Researchers from the Massachusetts Institute of Technology (MIT) have developed an algorithmic framework to automatically identify optimal molecular candidates to streamline drug discovery.
Published in Nature Computational Science, the quantitative framework, Synthesis Planning and Rewards-based Route Optimization Workflow (SPARROW), identifies the best molecules to test as potential new medicines while minimising synthetic cost.
Deployment of machine learning models to help identify molecules, among billions, that have properties to develop new medicines can be costly and time consuming.
Not only does SPARROW automatically identify optimal molecular candidates that have the most potential, but it also identifies the materials and experimental steps needed to synthesise them while considering the cost of synthesising multiple molecules at once, as multiple candidates can be derived from some of the same chemical compounds.
In addition, the approach captures crucial information on molecular design, property prediction and synthesis planning from online repositories and widely used artificial intelligence (AI) tools.
When evaluated across three case studies based on real-world problems faced by chemists, SPARROW effectively captured the marginal costs of batch synthesis and identified common experimental steps and intermediate chemicals, as well as scaling up hundreds of potential molecular candidates.
SPARROW collects information on the molecules and their synthetic pathways, weighs the value of each one against the cost of synthesising a batch of candidates and automatically selects the best subset that meets the user’s criteria by finding the most cost-effective synthetic routes for these compounds.
Not only could the algorithm help pharmaceutical companies discover new drugs more efficiently, but it could also be used in applications such as the invention of new agrichemicals or the discovery of specialised materials for organic electronics.
MIT’s lead author on the study, Jenna Fromer, commented: “By creating SPARROW, hopefully we can guide other researchers to think about compound down selection using their own cost and utility functions.”
John Chodera, computational chemist, Memorial Sloan Kettering Cancer Center, said: “The… approach… [balances time and cost while providing new useful information] in an effective and automated way, providing a useful tool for human medicinal chemistry teams and taking important steps towards fully autonomous approaches to drug discovery.”