Google's DeepMind is using AI to help scientists understand coronavirus

Google's DeepMind is putting its artificial intelligence systems to a new task: trying to figure out certain properties of the novel coronavirus which has killed thousands in the past couple of months.

In a post Thursday, DeepMind (which was acquired by Google in 2014 and is now a subsidiary of Alphabet), said it has put its AlphaFold system to create "structure predictions of several under-studied proteins associated with SARS-CoV-2, the virus that causes COVID-19."

These predictions haven't been experimentally verified, DeepMind says, but they may help scientists understand how the coronavirus functions. This, in turn, may be of use when developing a vaccine or cure.

DeepMind says that understanding a protein's structure typically takes months or longer. Prior knowledge of similar protein structures is also key. But AlphaFold is using newly-developed methods to provide "accurate predictions of the protein structure" with no prior knowledge, the company claims.

DeepMind also points out that normally, it'd wait until the work was peer-reviewed before publishing, but due to the nature of the coronavirus outbreak, the company is publishing the data now under an open license, meaning that anyone can use it. The data itself, along with more technical details, is available here.

Google's DeepMind made headlines in 2017, when it achieved a seemingly impossible task in beating human experts in the complex strategy game of Go. The company since put its AI systems to work on other challenges, including improving breast cancer screening process.

The coronavirus, officially named COVID-19, has so far killed more than 3,000 and infected more than 88,000 people.