Back at the start of April, a team of scientists and coders set up the mini-game Borderlands Science inside the world of Borderlands 3 in the hope that the game’s huge player base would lend their brainpower to identifying the DNA of microbes in the human microbiome – the collection of tiny organisms that live in the body and influence our health.
In the space of one month, the game’s developer Gearbox has seen some 700,000 players become citizen scientists and take part in the mini-game that analyses snippets of DNA harvested from the gut via poo samples.
The players didn’t need any medical knowledge (or rubber gloves) to get involved. Different building blocks of DNA, called nucleotides, from pairs of bacteria, were converted into different coloured squares and arranged into columns.
The goal for players was to align the coloured squares of each column with the ones next to it, by moving them up and down, while introducing as few gaps (empty squares) as possible.
Players line up rows of similar colours by shifting columns of blocks. Each block symbolises a nucleotide of bacterial DNA. More matches = bigger score.
In effect, this matching game tells scientists how similar the DNA sequence of bacteria A might be to bacteria B. If lots of blocks line up neatly in rows, then A and B share a similar ancestry and function. And if bacteria B is unidentified, but is a close relative of A, then we now understand a lot more about B.
In the game, the target score was the score achieved by an algorithm trying to carry out the same task, so the gamers were challenged with toppling it to earn in-game rewards.
It’s the kind of task that artificial intelligence typically struggles with, because there are goals at odds with one another (maximise alignments and minimise gaps) so the gamers would typically beat the score and some would go on set their own high score.
Each outcome will help scientists, who are working on the project at McGill University and the Microsetta Initiative, to develop different ways to solve the problem and ultimately train an AI to do the job as well as humans can.
It’s early days, but the results are already promising. It seems that on average 60 per cent of the solutions were where the computer put them, and 40 per cent were positioned better, thus improving the accuracy of their identification process.
Borderlands 3 players have now completed 36 million puzzles and when they hit 40 million the team will move on to bigger more complicated sequences in Phase 2.
The ultimate goal of all this puzzle-solving is to identify the bacteria living inside us, and their functions. The make-up of our microbiome has been linked to many diseases and conditions, including obesity, depression, cancer and more. If we can understand this ecosystem living inside us, then we can come up with better treatments and interventions.
The data is all open source so that future teams can benefit from the findings.
Daniel Bennett is the Editor of BBC Science Focus. He is an award-winning journalist who’s been reporting on science and technology for over a decade, writing about the science of serials killers, sandwiches, supernovae and almost everything in between.