Khatri was intrigued by the notion of engaging the public to solve problems. His lab develops novel diagnostics using publicly available data sets. The team had just published a paper on a set of genes that could help diagnose sepsis and had other papers under review on influenza and TB.
In an "Aha!" moment during their dinner chat, Khatri says, he and Das realized "how awesome it would be to sequentially merge our two approaches — to use public data to find a diagnostic marker for a disease, and then use the public's help to develop the test."
TB seemed opportune as it has a simple diagnostic signature — a set of three human genes that turn up or down predictably after TB infection. When checked across gene data on thousands of blood samples from 14 groups of people around the globe, the behavior of the three-gene set readily identified people with active TB, distinguishing them from individuals who had latent TB or other diseases.
Those findings, published in February, have gotten serious attention — not only from curious patients and doctors but also from humanitarian groups eager to help bring a better TB test to market. It can currently take several tests to tell whether a person has active TB, including a chest X-ray and sputum test. The Bill & Melinda Gates Foundation has started sending data to help the Stanford team validate a test based on the newly identified TB gene signature, says study leader Khatri, who works at the university's Center for Biomedical Informatics Research.
The test under development would use a common method called polymerase chain reaction (PCR) to probe samples for specific messenger RNAs — single-stranded molecules that carry portions of the genes' DNA code to parts of the cell where proteins are made. PCR is a mainstay of modern laboratories, but it requires machines that aren't always available in rural areas or developing countries.
Das thinks EteRNA players could help create a simpler TB test. The test would need a "switch" molecule — one that switches its shape when it senses that concentrations of the three signature RNAs are just so. If such a molecule existed, Das says, diagnosing TB could be as easy as reading a home pregnancy kit, in which sensor molecules embedded on test strips react with a signature hormone to produce colored bands indicating a positive result.
The latest EteRNA puzzles have prepared players for the TB challenge by teaching them how to create RNA molecules that switch their folding patterns in the presence of other chemicals or molecules. Earlier EteRNA puzzles didn't have the switch component; they simply asked players to arrange RNA's four building blocks into sequences to create strands that fold into desired shapes. Before the game went public in early 2011, EteRNA started as an RNA hack of Foldit, a crowdsourcing project that leverages similar problem-solving skills to predict how proteins are folded.
But in one regard EteRNA is fundamentally different. Whereas in Foldit players' predictions were compared against computer simulations, EteRNA puzzle solutions are scored by real molecules. Players vote on each batch of RNA designs, and winning structures get synthesized in a Stanford biochemistry lab. Insights gained from "wet lab" experiments with the synthesized molecules help players design more complex molecules in subsequent puzzles. EteRNA was conceived as an "Internet-scale citizen science project that includes actual experiments as part of the game play," Das says. The game is accessible at www.eternagame.org.
The game has attracted more than 100,000 players so far. "You can start out not knowing any of the science," says Jeff Anderson-Lee, a computer systems manager at UC Berkeley who got hooked on EteRNA 4 1/2 years ago. In early rounds, EteRNA guides players through small tasks that teach the rules of the game. "You start to get a sense for what things go together to make an RNA design fold the way you want," Anderson-Lee says. "While you're doing that, you're gaining points and rising in rank."
Games like EteRNA and Foldit show just how far citizen science has come. Foldit made a splash in 2011 when nonexpert players took just several weeks to solve the structure of an enzyme important for replication of an AIDS-like virus — a problem that had stumped scientists for the previous decade. As for EteRNA players, they aren't just glued to their screens solving puzzles over lunch. They're diving into the RNA literature and pushing science forward by forming and testing their own hypotheses, even organizing their own conferences.
In February, Anderson-Lee and other top gamers published a report in the Journal of Molecular Biology — which many consider the first peer-reviewed publication of research entirely conceived and carried out by citizen scientists. In another crowdsourcing victory published last month, nonexperts who played a quantum mechanics game developed strategies that outperformed scientists' computer algorithms. And this week, EteRNA Medicine will engage citizen scientists with its first puzzle aimed at a specific disease.
Now wait, you might be thinking, can't computers be trained to do this stuff? "On some level, the answer is likely yes," says Adrien Treuille, a computer scientist at Carnegie Mellon University who worked with Das to develop EteRNA. "We should be able to train computers to solve RNA puzzles, and they should be able to do it more quickly and with fewer errors."
But here's the rub: Artificial intelligence may churn out an answer but won't tell you how it got there. Treuille says "the crux of the human aspect of science is looking at data and coming up with understandable explanations for it — and asking new questions you didn't initially think to ask."