Episode 36: Unearthing Tomorrow's Medicine

In this episode we spoke to Jackie Hunter, CEO of Benevolent Bio, a company that utilises machine learning and AI to find previously overlooked drug candidates within the research literature. Jackie was previously Chief Executive of the BBSRC and comes into the AI space with a wealth of experience in industrial drug discovery.

Jackie strongly believes that AI can streamline science in a number of ways. For one, the system they are developing analyses the entire (accessible) research corpus, finds meaningful connections, and develops hypotheses relevant to the process of drug discovery. The consequence of this is that machine learning is capable of navigating away from an individual researcher's preferred area of reading. This opens up a whole new universe of potential findings. After all, perhaps the answer to your research question may just be sitting in a study from another field, and in a journal you may never have considered looking at.

The intriguing idea that hypotheses could be constructed from the relationships derived from the knowledge map, rather than by a singular researcher, essentially means that the evidence drives the hypothesis, and is not subject to the limited purview of the scientist. The drug discovery scientist will then validate these AI derived hypotheses and can probe further. 

Underlying this idea is the observation that while science is of course an evidence driven endeavour, the sheer volume of that evidence can mean that keeping on top of the literature can slow the march of research progress.

While what the team are doing represents an impressive advance in drug discovery, we were keen to get Jackie's insight into the challenges they face, including the accessibility of the research, and the frustrating issue of Publication Bias. Their main hurdle however, was the development of a rigorous and comprehensive, biomedical library of which the machine learning algorithms utilise in their digestion of the research literature. As anyone that's ever picked up a scientific paper knows, the language is dense, specific, full of acronyms, synonyms, homonyms (basically there's a problem with the amount of 'nyms'). And that's not even mentioning non-English research papers!

We had a blast taking a dive into the future of drug discovery, and we hope you enjoy the show. 

Professor Jackie Hunter of BenevolentBio chats with Sarah Buhr about using AI to trawl the global ocean of scientific and medical data for pharmaceutical R&D.