The February 2023 issue of IEEE Spectrum is here!

Close bar

COVID Moonshot: Can AI Algorithms and Volunteer Chemists Design a Knockout Antiviral?

This pro bono initiative crowdsourced 4,500 drug designs, synthesized 311, and is now testing them against viral proteins

3 min read
Colorful chemical fragments bind to the main protease of the SARS-CoV-2 virus.
Colorful chemical fragments bind to the main protease of the SARS-CoV-2 virus.
Image: Diamond Light Source

It started with a tweet. Alpha Lee, cofounder and chief scientific officer of machine-learning company PostEra, read on Twitter that Diamond Light Source, the United Kingdom’s national synchrotron facility, had identified a set of chemical fragments that attach to an important coronavirus protein.

Lee wondered if his company, formed just six months earlier, could help connect the dots from fragments to viable drugs to fight COVID-19. PostEra uses AI algorithms to map routes for drug synthesis to speed the drug-discovery process. But to do so, it would need some design ideas. So Lee asked the Internet.

On 17 March, in collaboration with Diamond, the PostEra team launched the COVID Moonshot to crowdsource drug designs from medicinal chemists. Then PostEra applied their technology, pro bono, to determine if and how those designs could be made.

“We thought we might have 100 or 200 submissions,” says Lee, an associate professor at the University of Cambridge. “In fact, we got thousands.”

Over 4,500 molecular designs from 280 contributors around the world flooded into the submissions site PostEra set up for the effort. Two chemical synthesis companies have stepped up to physically make the compounds, providing their services for free or at reduced cost, and two pharmaceutical companies, UCB and Boehringer Ingelheim, are contributing employee time toward the effort at no charge.

“There’s an element of ‘we should do something’ from the PIs of the project, almost a duty, not wanting to leave key scientific equipment and great minds idle, and it has caught on” says John Spencer, a professor of bioorganic chemistry at the University of Sussex, who worked for 10 years as an industry medicinal chemist and is volunteering in the Moonshot. In addition to being involved in discussions and providing advice, Spencer has submitted around 100 ideas and sent compounds from his university laboratory for testing.

If and when any drug candidates are identified via the crowdsourced project, the drug designs will be made openly available in the public domain without patent or any intellectual property restrictions. “In times like these, when the world is closing down, science should open up,” says Lee. “We are really optimistic that we can get a viable [drug] candidate out of this effort.”

Currently, the chemical fragments discovered at Diamond are a far cry from actual drugs. The small molecules only weakly attach to the active site of a key coronavirus protein. A true drug compound requires additional chemical components to be potent, safe, and lasting in the body.

Design ideas for such compounds are being submitted by academics, students, retirees, industry medicinal chemists, and more. “There are a lot of like-minded experts using a diversity of tools: some expert intuition, others machine learning, and others physical modeling,” says Lee. 

PostEra's machine learning algorithms generate synthesis routes to make chemical compounds that bind a key coronavirus protein.PostEra’s machine_learning algorithms generate synthesis routes to make chemical compounds.Image: PostEra

The PostEra team runs the designs through their machine-learning pipeline—algorithms trained on over 10 million chemical reactions scraped from patents of existing chemicals—to triage which designs can be made, and then generate recipes to do so rapidly. In a 2019 paper, PostEra’s algorithms outperformed human chemists in predicting the outcomes of chemical reactions.

Lee estimates that designing ways to synthesize over 2,000 molecules might take chemists about three weeks. The PostEra algorithms did it in a weekend.

Modelling of fragment hits into electron density maps obtained from x-ray diffraction CoV-2 Mpro protein crystals. Modeling of a chemical fragment (purple) binding to a coronavirus protein.Image: Diamond Light Source

Once the first batch of designs was triaged and complete, PostEra sent them off to chemical synthesis companies Enamine and Sai Life Sciences, which synthesized the compounds at no or reduced cost. “They’ve been super generous with their time,” says Lee. Incurred synthesis and testing costs are being funded through a GoFundMe campaign

Next, laboratories at the University of Oxford, in England, and Weizmann Institute, in Israel, began testing the compounds against the coronavirus protein, an enzyme called Mpro that is central to the virus’s ability to replicate. So far, there are several promising leads, says Lee. After testing against the protein in a dish, any strong hits will move into being tested against the whole virus, then against viral infection in animals. Lee hopes to identify a preclinical candidate in the next few months.

“Knowing there’s a chance we’ll stumble upon some new discoveries, in an unprecedented manner, with a number of scientists from all countries, all levels—everyone is welcome, everyone has a voice—is invigorating,” says Spencer. “I sincerely hope that this is a sign of things to come.”

If you want to get in on the effort, the Moonshot team continues to welcome design submissions through their website.

The Conversation (0)

How Duolingo’s AI Learns What You Need to Learn

The AI that powers the language-learning app today could disrupt education tomorrow

9 min read
Vertical
This playful illustration shows Duolingo’s owl mascot, cut away down the midline, showing hidden inside a high-tech skeleton suggestive of some sort of AI robot.
Eddie Guy
Blue

It’s lunchtime when your phone pings you with a green owl who cheerily reminds you to “Keep Duo Happy!” It’s a nudge from Duolingo, the popular language-learning app, whose algorithms know you’re most likely to do your 5 minutes of Spanish practice at this time of day. The app chooses its notification words based on what has worked for you in the past and the specifics of your recent achievements, adding a dash of attention-catching novelty. When you open the app, the lesson that’s queued up is calibrated for your skill level, and it includes a review of some words and concepts you flubbed during your last session.

Duolingo, with its gamelike approach and cast of bright cartoon characters, presents a simple user interface to guide learners through a curriculum that leads to language proficiency, or even fluency. But behind the scenes, sophisticated artificial-intelligence (AI) systems are at work. One system in particular, called Birdbrain, is continuously improving the learner’s experience with algorithms based on decades of research in educational psychology, combined with recent advances in machine learning. But from the learner’s perspective, it simply feels as though the green owl is getting better and better at personalizing lessons.

Keep Reading ↓Show less
{"imageShortcodeIds":[]}