AI: Kickstarting the Race Against COVID-19

Julian Upton

By the time the devastating potential of COVID-19 was really beginning to shake the governments of countries outside China, many international AI drug discovery companies were ramping up efforts to identify existing drugs and new molecules that could be used to treat the virus. At the beginning of February, South Korean company Deargen reported that its deep learning-based model MT-DTI, for example, had revealed the potential of the

HIV medication atazanavir to bind and block a prominent protein on the outside of SARS-CoV-2, the virus strain that causes COVID-19.

At the same time, Hong Kong-based Insilico Medicine, using AI algorithms to design new molecules that could limit the virus’s ability to replicate, published the structures of six potentially effective molecules. In the UK, startup BenevolentAI published a short article in The Lancet that highlighted baricitinib (marketed as Olumiant in the US), a once-daily pill developed by Eli Lilly and Incyte and approved for the treatment of rheumatoid arthritis (RA), as the most likely of six compounds that it had identified to potentially block the cellular pathway that allows the coronavirus into cells to make more virus particles.

While worldwide AI efforts to identify treatments are ongoing, BenevolentAI’s suggested treatment, baricitinib, moved to the next level on April 11 when Lilly announced that the drug had entered a randomized-controlled trial in COVID-19 patients. “Lilly is moving at top speed and using all available resources to help fight this pandemic,” commented Daniel Skovronsky, Lilly’s chief scientific officer. Results of randomized-controlled trials of baricitinib for the treatment of COVID-19 in the US, Canada, Denmark, and Spain are expected in the next few months.

BenevolentAI’s focus is usually on identifying potential new therapies, but the COVID-19 outbreak saw it concentrate on established medicines that can be quickly moved into clinical trials, Peter Richardson, the company’s vice president of pharmacology, told Pharm Exec. He added that “life science and technology companies have a duty to mobilize resources for the public good in the current global health emergency.”

With powerful platforms at their disposal, AI companies are able to act with an urgency that leaves other drug discovery companies looking inert. It took around three days, for example, for BenevolentAI to identify baricitinib as a potential COVID-19 treatment option.

The company set up a team to deploy its tools, models, and algorithms to explore the biomedical “Knowledge Graph” that it has curated over the last five years, and which draws on “vast quantities of mined and inferred biomedical data,” to assess treatment options, Richardson explains. Data was pulled from various structured and unstructured biomedical sources, curated, and then fed into the Knowledge Graph, which further contextualized the relevant omics data, clinical data, and scientific literature.

With the Knowledge Graph “rich in mammalian biology,” BenevolentAI’s scientists looked for ways to inhibit the cellular process that the coronavirus uses to infect human cells-the process of endocytosis, whereby the virus attaches to the surface of a cell and is then absorbed inside-rather than focus on drugs that could inhibit the virus directly. Although baricitinib is currently approved for use as an RA treatment, BenevolentAI’s studies suggest that it could slow down endocytosis and reduce the ability of the virus to infect lung cells. 

Of course, at this point in the crisis trajectory, AI’s power is being harnessed to crunch the data and alert drug companies and healthcare systems to potential treatments, rather than being posited as a “solution” to the overall problem. Indeed, in its Lancet paper, BenevolentAI cautioned that its findings should not be taken as medical advice about COVID-19 treatment or prevention, but were offered “to assist in the global response” to the virus.

It now of course falls on time-honored clinical research methods to establish the safety and efficacy of any particular drug in treating COVID-19 patients (it has been pointed out that baricitinib already carries a warning that the drug could actually make infections worse). But whether or not the new trials achieve the desired results, without AI, says Richardson, we wouldn’t know how long it would have taken to identify baricitinib as a contender in the battle against the virus (or, indeed, if it would’ve been identified at all). One thing we can be sure of is that AI “significantly sped up the process.”

Julian Upton is Pharm Exec’s European and Online Editor. He can be reached at jupton@mjhlifesciences.com