Canadian scientists win award for enlisting gamers to identify COVID-19 in blood data

Partnership with Icelandic company crowdsources citizen science by allowing players to earn rewards by helping analyze data

In the world of artificial intelligence, a computer program that can reliably spot signs of disease in a blood sample is a feat that borders on science fiction.

So when Ryan Brinkman, a bioinformatics scientist at the BC Cancer Agency, wanted to develop such a program, he called on an army of online gamers who populate a digital sci-fi universe.

Last week, Dr. Brinkman and Jérôme Waldispuhl, a computational biologist at McGill University in Montreal, were recognized for their efforts by sharing in a Webby award, the internet’s highest accolade, together with partners at Icelandic game developer CCP Games.

The award – in the category of public service, activism and social impact – was for developing Project Discovery, an app in CCP’s multiplayer game, EVE Online. For the past year, the app has allowed more than 325,000 players to earn game rewards by taking time out from piloting virtual spaceships to categorize blood samples from patients with COVID-19, among other diseases.

The point of the exercise is to build up a huge pool of results that can be used to train software. The approach, known as machine learning, has previously allowed computers to conquer tasks such as speech recognition. It is increasingly making inroads in medical diagnostics.

Dr. Brinkman said Project Discovery was made possible by the COVID-19 pandemic and the high volume of blood tests that came with it. But his end goal is a system that can help direct treatment for a range of ailments, from HIV to leukemia.

Such diseases can be detected through flow cytometry, a diagnostic test that involves passing blood cells through a laser beam. Cell proteins are measured as they interact with the laser light.

The relative abundance of various proteins can reveal the presence and stage of a disease.

Dr. Brinkman’s lab has developed computer algorithms to handle flow cytometry readouts, but he said that sifting through the data is a laborious process that can take months for one disease. “We’d do so much better if [the algorithms] worked faster,” he said.

Machine learning offers a way to automate the process, but only if a learning algorithm can train itself by working through reams of flow cytometry data. Until recently, such data were hard to come by. That changed during the pandemic, when Dr. Brinkman’s lab, which handles flow cytometry for researchers studying a range of diseases, was able to access a flood of samples related to COVID-19.

Even then, the data still needed to be processed by humans so that a computer could use it to check and adjust its own performance. It was this final barrier that led Dr. Brinkman to the crowdsourcing potential of thousands of online gamers.

He teamed up with Dr. Waldispuhl, who leads DNA Puzzles, a lab project that uses video games to crowdsource citizen science.

“My job is really about developing methods to accelerate research and enable biologists and geneticists to do things they couldn’t do before,” Dr. Waldispuhl said.

Through contacts made during previous collaborations with the gaming world, Dr. Waldispuhl learned that CCP was looking to host a citizen science project in its sci-fi game EVE Online.

“We started the conversation … to see if there was any possibility of doing something to support COVID-19 research,” said Bergur Finnbogason, the Reykjavik-based creative director at CCP.

The game, which celebrated its 18th anniversary this month, has a history of contributing to citizen science, including projects that had players classifying proteins or searching for signs of real-life planets buried in astronomical data.

David Ecker, lead producer for CCP, said the biggest hurdle was figuring out how to teach players to deal with the complex data that Dr. Brinkman and Dr. Waldispuhl needed them to handle. It was a case of, “How do we translate this to your average Joe?” he said. “How do you explain flow cytometry?”

The team produced a tutorial showing players how to draw circles around clusters of flow cytometry data, just as scientists would do. As players progressed, their results could be compared with reference samples to gauge their accuracy.

Since Project Discovery was launched in June, 2020, EVE players have analyzed more than 1.25 million pieces of data.

“It’s a perfect example of a novel approach that, if executed successfully, [can save] huge amounts of time and resources to produce quality training data,” said Alex Dobranowski, physician and chief executive officer of Toronto-based MCI Onehealth, which has been developing and incorporating artificial intelligence into the diagnostic services the company offers as part of its primary-care network.

Dr. Brinkman said the gamers’ performance with COVID-19 data has been so impressive that the team has shifted them to flow cytometry readouts from other diseases. The next step is to utilize that data to hone what Dr. Brinkman calls “the magic unicorn” of machine learning: an algorithm that can analyze a blood sample and pinpoint any disease.

“It does look like it’s going to work,” Dr. Brinkman said. “There’s more work to do. But it’s going to make an impact.”