Running on AWS, BenevolentAI Uses Its Knowledge Graph to Facilitate New Drug Discovery

Developing new drugs is typically complicated, time consuming, and costly.

“We exist as a company in order to make that process faster so that we can get more treatments out to patients,” says Olly Oechsle, a lead software engineer at BenevolentAI. The company is working to accelerate the development of new drugs by pairing its knowledge graph and artificial intelligence models with scientific understanding. “Our knowledge graph is a collection of millions of entities and hundreds of millions of relationships that we have gathered from literature and other sources and processed on the Amazon Web Services [AWS] platform. It is composed of diseases, of drugs, of proteins and genes, of processes, clinical trials, and all of the different entities that are important for drug discovery,” says Oechsle.

In the case of the COVID-19 pandemic, BenevolentAI was able to pivot from its typical work and rapidly use its knowledge graph to identify an existing approved drug that may be repurposed as effective in treating COVID-19. “Being able to run on a scalable platform like AWS allows us to do experiments that we couldn’t otherwise do. It allows us to change the assumptions that we make, and it allows us to do all of that without jeopardizing our day-to-day production work,” says Oechsle. As a result, BenevolentAI identified baricitinib as a potential drug for COVID-19 treatment with only 90 minutes of AWS-powered compute time and less than 3 days of additional human work. Within a month the drug manufacturer announced it would begin its first large randomized trial in COVID-19 patients.