Called the ‘Coronavirus Deep Learning Challenge‘, this takes the form of a two-week challenge where teams compete by using artificial intelligence in order to find a potential treatment for COVID-2019. The objective is to find a candidate drug (ligand) with a high binding affinity with the COVID-2019 main protease. After the exercise is completed, the compound with the most potential will be donated to the Wuhan Institute of Virology for further analysis.
To learn more, Digital Journal spoke with Siraj Raval of Sage Health,, who came up with the idea of the deep learning challenge along with his collaborator Dr. John Billings.
Digital Journal: How serious is Covid-19, in terms of infectivity and a lack of medicines?
Siraj Raval: COVID-19 is turning out to be one of the deadliest viral disease outbreaks of all time, with over 40K confirmed cases as of writing this. It can be caught through coughing, sneezing, or any form of human contact. Some virtual screening efforts have shown that HIV antivirals could offer promising results to be potential candidates, but more research needs to be done.
DJ: What is the idea behind the ‘Coronavirus Deep Learning Challenge’?
Raval: In the AI community, the “Deep Blue” moment was a moment that shocked the world when a computer first beat a human at Chess. It was the first time an AI had surpassed human capability in a way that was relevant to the general public. AI applied to drug discovery, will result in the first AI-Human Deep Blue moment, a moment where, augmented with AI, a human or group of humans, is capable of doing something they couldn’t otherwise and have a massive impact in the world. Using Deep Learning for Coronavirus Drug Discovery is the most important, relevant application of AI right now.
The motivation behind the competition is to show the world how community-driven, open-source AI can make a big, positive difference. In 2 weeks, teams from all over the world will participate by using deep learning to generate potential drugs and scoring each of their potentials as candidate for 2019-COVID. The three winning teams will receive $3500 in prizes and will be announced via social media. The competition is hosted in collaboration with Insilico Medicine for research ideas.
DJ: What are the benefits of a deep learning approach?
Raval:Deep Learning offers much more of a powerful search tool through the very wide amount of potential candidate structures. Because Deep Learning implies “big data” and “big compute” drug discovery is inherently a perfect choice to apply it to, since there are millions of potential drugs and processing all of their bio-complexity will require lots of computing power.
DJ: How did you develop the open source technology?
Raval:This technology was developed by several contributors from across the world and posted on Github.com. you can find a list of possible libraries here.
DJ: How was the ‘Coronavirus Deep Learning Challenge’ developed and tested?
Raval:It was developed by analyzing what different researchers in the field are doing with AI to find a cure. After analysis, it was found no one was doing an open source competition, engaging the public, so this idea occurred. It’s only been live for 30 minutes and already 13 teams have signed up!
DJ: How long does the challenge last for and what are the incentives?
Raval:The challenge runs for two weeks, with $3500 in prizes: 1st place: $1000 cash + $500 JetML cloud Credits; 2nd place: $1000 JetML cloud credits; and 3rd place: $1000 JetML cloud credits.
DJ: What will happen once the winning compound is discovered?
Raval:We will donate samples of it to the Wuhan Institute of Virology
DJ: How do you think deep learning / AI will develop to promote drug discovery for other areas of medicine?
Raval:It will continually increase in its potential to understand and interpret biomolecular complexity, eventually finding applications in quantum chemistry simulations.
