Reading time
1 min read
Colorful chemical fragments bind to the main protease of the SARS-CoV-2 virus

Researchers are drawing up blueprints for drugs to fight Covid-19. Machine learning is identifying those most likely to be effective.

What’s new: Covid Moonshot, an international group of scientists in academia and industry, is crowdsourcing designs for molecules with potential to thwart the coronavirus. The project is using a deep learning platform to decide which to synthesize for testing. Any intellectual property it develops will be donated to the public domain.

How it works: The group began in March as a partnership between PostEra, a UK-based startup, and Diamond Light Source, a British government science lab. PostEra issued a call for submissions of compounds that incorporate specific chemical fragments that bind to a protein the virus uses to replicate, as pictured above. It has received over 4,500 proposals so far.

  • PostEra’s semi-supervised learning system models chemical reactions to determine which compounds are practical to manufacture.
  • Designs that pass this analysis are sent to one of two drug manufacturers that have agreed to produce the substances at minimal cost.
  • These companies send them to university labs for testing.
  • Molecules that prove successful against Covid-19 in a petri dish will move to preclinical trials with lab animals. PostEra hopes to begin this stage within the next few months.

Results: The organization’s manufacturing partners have synthesized over 700 compounds, of which nearly a third have been tested in the lab. Twenty-eight of these inhibited the virus, and eight were especially potent.

Why it matters: This pandemic doesn’t appear to be going away any time soon. AI that predicts the most viable treatments could help limit the damage.

We’re thinking: When you combine citizen science with AI, amazing things can happen.


Subscribe to The Batch

Stay updated with weekly AI News and Insights delivered to your inbox