Tapping artificial intelligence and machine learning to reinvent the ways drugs are discovered
CEO: Joanna Shields
Based: London, U.K. and Cambridge, Massachusetts
The scoop: Machine learning and artificial intelligence have been touted as the future of drug discovery for years, but BenevolentAI is working to make the dream a real-time reality.
What makes BenevolentAI fierce: For many life science companies, the coronavirus pandemic hit hard, especially in traditional clinical trials where sites were heavily disrupted by lockdowns. This knocked R&D efforts back by months and even years.
For BenevolentAI, however, COVID became a proving ground for its swift, AI-based approaches to drug discovery, with the U.K.-U.S. company prompting Eli Lilly to run trials for its arthritis med Olumiant (baricitinib) during COVID-19.
Unlike so many other repurposed drugs in testing for the disease, Olumiant wasn’t going in blind: In a study published in The Lancet early last year, the BenevolentAI team suggested the Lilly drug could not only reduce inflammation but also block the virus from entering and infecting lung cells by inhibiting a protein called AAK1.
This spurred Lilly to kickstart a phase 3 trial in the summer, and by October the drug—in combination with Gilead’s repurposed Ebola med remdesivir—showed it could reduce time to recovery and improve clinical outcomes for patients with COVID-19 infection when compared to remdesivir on its own. The drug combo was swiftly given its own emergency authorization by the FDA for certain hospitalized patients.
That was a pivotal moment for BenevolentAI and the drug fight against COVID-19, which has seen many casualties over the past year.
“In January 2020, with thousands of patients already critically ill, a small BenevolentAI team traversed our Knowledge Graph to identify approved drugs, already developed and proven safe, that could be repurposed to treat the virus,” explained BenevolentAI’s Ivan Griffin, COO and co-founder of the company.
Over a single weekend, they found that its arthritis drug baricitinib was the one to go for. This was due to the drug’s well-known anti-inflammatory effects, plus some previously unknown antiviral effects predicted by BAI's models. Finding were promptly published, and by April, the National Institute of Allergies and Infectious Diseases (NIAID) and Lilly began a major test of around 1,000 patients.
The NIAID data proved the hypothesis: It improved clinical outcomes, increased recovery rates and reduced mortality among hospitalized COVID-19 patients, especially those on supplemental oxygen, a boon for Lilly, patients and of course confirming BAI's hypothesis.
This tech was further validated earlier this year, but in a very different context: The firm, which also has a pact with AstraZeneca, saw its so-called Benevolent Platform predict a new chronic kidney disease target from its AI tech and was the first AI-generated drug from the project to enter their portfolio.
Once again, the computer-assisted hypothesis was biologically validated through AstraZeneca’s experimental testing. “The combination of our proprietary technology with AstraZeneca’s scientific expertise is helping to steadily close the gap between AI, data and biology,” Griffin said. “I believe that this CKD target milestone is only the beginning of what our partnership is able to achieve.”
AI as a tool, not a replacement, for scientists
But whilst the uses of AI and machine learning have been shown to speed up drug discovery and help find new targets, it should still be used as a tool, and not a complete methodology. “The role of AI, be it in designing new drugs or repurposing old ones, in pandemic response or drug discovery more broadly, should be to augment scientists’ capabilities, not replace them,” explained Griffin.
“Scientists play essential roles in determining the data to use and in expertly evaluating the results, both for additional accuracy and nuance. In our COVID-19 research, AI played a significant role in accelerating the search for potential drug candidates, streamlining triage and enhancing the ability to query these results.”
“However, it was experienced scientists who evaluated those recommendations and put forward the hypothesis. Ultimately, this fusion of machine and human intelligence holds the key to unleashing the full potential of new technologies in drug discovery and development.”
Working through a pandemic
The company, which has sites in the U.K. and the U.S., has managed to work through the constraints of the pandemic and continue with its own internal candidates. “With sites in London, Cambridge [Massachusetts] and New York, we were already well versed in video conferencing and remote working,” said Griffin.
“For our lab-based scientists, the closure of our lab at the start of the pandemic was certainly frustrating, but we adapted and reopened for a reduced team of scientists to resume experiments when the government deemed it safe. As a result, we continue to deliver on our key objectives, support our research partners and successfully maintain our in-house portfolio of drug programs.”
This includes the trial of BEN-2293, a molecule designed and developed by Benevolent scientists to treat atopic dermatitis, for which it recently dosed the first patient in the randomized first-in-human test.
Investors: Once part of the now fallen Neil Woodford investment, BenevolentAI bounced back from the Woodford Equity Income fund collapse, nabbing a $90 million cash injection from Singaporean sovereign fund Temasek last year, coming two years after it raised $115 million from a number of undisclosed American investors.