You are here
Gaming with purpose
Sometimes scientific progress occurs when two different streams of thought unconsciously collide.
Since 2003, Mathieu Blanchette of McGill University’s School of Computer Science has been studying how to program computers so that they understand what controls the expression of genes. Through his research, he realized that despite a computer’s ability to track millions of possible data scenarios, it often can’t visualize the best scenario.
“The main thing that differentiates the human brain from computers is the idea of intuition,” says Blanchette. “We can’t explain how the brain does what it does so well, especially when it comes to visual recognition of patterns.”
Blanchette’s research partner, Jérôme Waldispühl, a computational molecular biologist at McGill, spent considerable time while instructing at the Massachusetts Institute of Technology thinking about how computer gaming could be turned into something purposeful. “I’m not a gamer myself,” he says, “but I did play sometimes for a break and kept thinking, ‘Can we reuse that energy to do something useful?’”
Since November 2010, this online game has enticed gamers around the world to use their brains’ visual intuition to help solve one of the knottiest problems faced by modern molecular biology: How do we make sense of the human genome? Part of the answer lies in the comparison of our genome to that of other species through a process called genome alignment. When the genome of one species is aligned with that of another, you can see matches.
Any matches in an alignment, says Blanchette, “tell which regions of the human genome have been highly conserved in mammal species, and that tells you they are very important for something.” He points out that a mutation in the genome suggests a genetic initiator of a disease or condition.
Phylo is based on the classic move-the-pieces-until-they-fit-together game of Tetris. A gamer must cross-match horizontal sections of human DNA that have been pre-aligned using computer algorithms. The sections come from 521 genes chosen because of their potential linkages to human diseases. But these algorithms are not exact and produce only approximate solutions. With Phylo, Blanchette and Waldispühl can identify and extract the regions (of smaller size) that are poorly aligned by the computer and ask gamers to improve them. Then they insert the best alignments calculated by the players into the original (global) alignment system.
The DNA sections aren’t defined by the typical chemical lettering but by blocks of four separate colours. Each colour represents a different nucleotide that, when combined with other nucleotides, creates a DNA segment. Thus a colour match equals a genetic match. Up to eight species can be compared in a single game.
The more colour matches, the higher the score. In addition, if gamers find linkages that nobody else has discovered, their scores go up. And gamers can choose to play with sequences related to diseases in which they are especially interested.
The fine-tuned results are posted on a McGill website.
How successful has Phylo been as a game? Since it was launched almost a year and a half ago, more than 600,000 games have been played by more than 20,000 registered gamers in over 120 countries.
And the results have been promising.
In a recently published paper, the McGill researchers point out that Phylo gamers have significantly increased their computer’s analytical powers. “In about two-thirds of the genomic regions aligned by Phylo gamers,” says Blanchette, “an improvement in the computer’s accuracy of alignment took place,” says Blanchette. The increased accuracy has yet to translate into a direct linkage between a disease-causing mutation or alteration and a genetic section common to many species. Nor has it created better algorithms to improve the computer’s alignment skills, but it has demonstrated a key first principle: When it comes to analyzing data, computers aren’t always better than humans.
And this leads to the hope that Phylo will help researchers find the Holy Grail of molecular biology. Refinements in aligning similar genetic fragments between species may one day enable scientists to reconstruct the genome of different species’ common ancestors — think the primordial dog — and determine how they evolved.
“In the end, the message is not of a competition but of humans and machines complementing one another,” says Waldispühl. Human visual intelligence can see things that heuristic-based computers miss. And with that comes great promise of a human/machine algorithm collaboration in the future.