By Benjamin Ross
September 14, 2018 | Researchers from KTH Royal Institute of Technology in Stockholm, Sweden, and Massive Multiplayer Online Science (MMOS) recently worked with CCP Games, using their multiplayer online game, EVE Online, to gain a more granular understanding of patterns of proteins arranged within the body’s cells.
The results from this effort were recently published in Nature Biotechnology (DOI:https://doi.org/10.1038/nbt.4225), detailing how researchers found that players, or “citizen scientists” as the article refers to them, helped boost the artificial intelligence system used for predicting protein localization on a subcellular level.
The researchers partnered with MMOS and CCP Games to integrate the analysis of protein localization from protein images directly into EVE Online. The resulting mini-game was played by more than 300,000 citizen scientists within EVE Online, generating more than 33 million image classifications of protein subcellular localization.
The ability of citizen scientists was compared to an Artificial Intelligence (AI) system for predicting protein subcellular localization from images called the Localization Cellular Annotation Tool (Loc-CAT). According to the researchers, Loc-CAT is the first generalized tool for annotating proteins with multiple localizations from images, and generalizes across a large number of cell types, providing a useful tool in studying cells and their behavior in the future.
The combination of crowdsourcing and AI led to improved classification of subcellular protein patterns and the first-time identification of ten new members of the family of cellular structures known as “Rods & Rings,” according to Emma Lundberg, a researcher from KTH who leads the Cell Atlas, part of the Human Protein Atlas, at the Science for Life joint research center.
“We wanted the gamers to classify and find new patterns, and compare the results side by side with the machine learning methods” Lundberg told Diagnostics World. “We can see that both the gamers and machine learning could contribute to the annotation of these images. We now have 30 locations classified as opposed to the original 23.”
Built on a map that shows hundreds of thousands of microscopic images of human cells, EVE Online players worked alongside an artificial intelligence to accomplish this goal.
“While humans still are the experts at spatial recognition. The gamers and the AI are not far behind. They are also good at different things, and by transferring the knowledge from the gamers we could build an improved machine learning model” Lundberg said.
Lundberg says the data is being actively integrated into the publicly-available Human Protein Atlas database and will be a resource for researchers worldwide who are working toward a greater understanding of human cells, proteins, and disease development.
“Ideally I would like to explore the transition from normal tissue to cancer,” Lundberg says. “But that would require a computer game with many more players. There’s a lot of hidden biomarkers we could potentially discover if we explore this further.”
Lundberg might get the chance to explore those potentials shortly, with researchers announcing the Human Protein Atlas 2018 Challenge on Kaggle starting in the end of September. The challenge will involve image analysis to classify subcellular protein patterns in human cells. Thousands of dollars of prizes will be up for grabs and the contributions of participants will help drive the field of protein biology.
“There are two goals [with this challenge],” Lundberg said. “The first goal is to obtain a model that can do better than the experts. Because if we have that model, I can take the researchers that are working on annotating those images and put them to work on something else, which would save us time and money. The other thing is that I want to raise awareness of this massive annotated image dataset, and promote use of it as a benchmark dataset.”
This is not the first time researchers have sought out the particular skillsets of gamers to help analyze proteins. Online puzzle games such as Foldit have been utilizing the pattern-detecting skills of gamers for years.
The reason more and more researchers are turning to these citizen scientists, Lundberg says, is because there’s still a great deal of uncertainty in the science community about how proteins work. “It’s an unexplored field still,” Lundberg said. “The genome was sequenced in 2001, but we still don’t have a complete understanding of what all the proteins that the genome encodes are doing.”
Lundberg sees the potential from tapping into the gamer community’s resources to be enormous, adding that she wouldn’t be surprised to see more efforts like this in the coming years.
“Video games can be viewed as portals into people’s brains,” she said. “And I think you can make use of it for both educational and medical purposes. I think we’ll see a lot more integrated efforts like this soon.”
Lundberg also made her impact in the project on a personal level, becoming the first ever scientist who was put into a videogame as a non-playable character to direct the project in-game.
“It was an honor of course,” she laughed. “I didn’t know they were going to make me an avatar. We were in a meeting talking about the design of the game and things like that, and at the end they said, ‘Oh by the way, we were thinking about introducing a character called Professor Lundberg!’ I’m certainly proud of that, especially to represent women in science.”