Phylo | Leaders in Pharmaceutical Business Intelligence (LPBI) Group

Posts Tagged ‘Phylo’

Gamification of Genomics and Proteomics Research

Posted in Bio Instrumentation in Experimental Life Sciences Research, Biological Networks, Gene Regulation and Evolution, Biomarkers & Medical Diagnostics, Computational Biology/Systems and Bioinformatics, Disease Biology, Small Molecules in Development of Therapeutic Drugs, Genome Biology, Genomic Testing: Methodology for Diagnosis, Metabolomics, Personalized and Precision Medicine & Genomic Research, Proteomics, Statistical Methods for Research Evaluation, tagged Foldit, Games, Gamification, JavaScript, McGill University, Phylo, Scripps Research Institute, University of Washington on September 3, 2012| 2 Comments »

Reporter: Aviva Lev-Ari, PhD, RN

Game On

September 2012

By Matthew Dublin

Combining gaming and genomics may sound odd, but play can produce useful data. By presenting complex biological problems as games and distributing those games far and wide, researchers can take advantage of a large network of virtual computation in the form of thousands of players. Games can also harness the natural human ability for pattern recognition and take advantage of the ways in which the human brain is better at that than even the most powerful supercomputer.

One of the first games to harness the power of the crowd was Foldit, which was developed by the University of Washington’s David Baker in 2008. Foldit encourages players to solve protein structure prediction problems by folding proteins into stable shapes.

Last September, an elite contingent of 15 Foldit players used molecular replacement to solve the crystal structure of a retroviral protease from the Mason-Pfizer Monkey Virus, which causes simian AIDS.

Baker’s team published a paper in Nature Structural & Molecular Biology describing how the solutions facilitated the identification of novel structural features that could provide a foundation for the design of new antiretroviral drugs. According to the authors, this marked the first time gamers solved a longstanding -biological problem.

Then in December 2010, a team from McGill University rolled out Phylo, a Sudoku-like game that utilizes players’ abilities to match visual patterns between regions of similarity in multiple sequence alignments. Phylo’s designers reported in a PLOS One paper published in March that, since the game’s launch, they had received more than 350,000 solutions produced by more than 12,000 registered players.

“We don’t know right now what other interesting problems we could solve using these crowdsourcing techniques. We still have a lack of deep understanding about when human intuition or help is very useful,” says Jérôme Waldispühl, an assistant professor at McGill University and lead developer of Phylo. “But what I like is the involvement of society digging into the most meaningful and deep scientific questions — you are trying to involve society into the very process of scientific discovery.”

Last year, a group from Carnegie Mellon University and Stanford University released an online game called EteRNA, the purpose of which is to help investigators envision RNA knots, polyhedra, and any other RNA shapes that have yet to be identified. Top designs are analyzed each week to determine if the molecules can fold themselves into the 3D shapes predicted by RNA modeling software.

Purposeful play

More “games with a purpose” — as they are sometimes called — aimed at solving biological problems are in development.

“No matter how big your super-computer is, you can’t try all gene combinations within a 20,000-gene space,” says Benjamin Good, a research associate at the Scripps Research Institute. “Humans have a role to play here because maybe we can do better than what happens when you just compute for just many, many random datasets.”

Good and his colleagues have developed a suite of games, called Gene Games, which they hope to use to build out genomic databases and to improve existing algorithms. This suite includes GenESP, a two-player game in which both players see the same disease and each must guess what gene the other is typing from a dropdown list of possible genes. This game is aimed at an audience with some knowledge of the field and takes advantage of players’ expertise.

Another game is Combo, which challenges players to find the ideal combination of genes for phenotype prediction. Combo players can choose to start at an introductory level where they separate mammals from other animals or divide the animal kingdom into five classes, or begin at a more challenging level where they have to identify gene expression signatures in tumors to predict survival or metastasis.

“The goal of issuing GenESP is to provide a new way of building out gene annotation databases, and Combo is specifically made to enrich a machine-learning algorithm. It’s all an attempt to use games to tap into a large communities of people to get after what they know,” Good says.

But he is quick to point out that there is no specific critical mass of players that will provide the desired data — instead, it is about finding the right players just as the Foldit project seems to have done.

“I can’t say when we hit 1,000 users or 10,000 users, we’ll have X units of compute available to us — it just doesn’t work like that. But if we end up getting the right 100 people playing these games, we can make a lot of progress,” Good adds.

Diagnostic disport

Whether taking a crowdsourcing approach to access networks of thousands of players or to get at a handful of skilled players, gaming can not only provide new data for researchers, but can also provide a handy diagnostic resource for clinicians.

In May, a group from the University of California, Los Angeles, released Molt, a game in which players aid in the diagnosis of malaria-infected red blood cells. After completing training, players are presented with frames of red blood cell images and use a virtual syringe tool to eliminate infected cells and then use a collect-all tool to designate the remaining cells in the frame as healthy. Results from the games are sent back to the point-of-care or hospital setting.

Popular online games are also providing models for new bioinformatics applications. A new software tool called ImageJS marries bioinformatics with pathology, which its developers say takes its cue from Rovio Entertainment’s popular Angry Birds game. Developed by a team at the University of Alabama at Birmingham, this free app allows pathologists to drag pathology slides into a Web app to identify malignancies based on color.

“There are two bioinformatics problems at the point of care that an Angry Birds-approach would solve. The first is that it delivers the code to the machine rather than forcing the data to travel, so the code does the traveling,” says Jonas Almeida, a professor at UAB. “The second problem it solves is that it doesn’t require installation of any application; they are written in JavaScript, which is a native language of Web development, and the application has no access to the local file system, so we have an application that does not make the IT people nervous.”

Almeida and his team have also designed a module for ImageJS that can analyze genomes using the same visual interface as the malignancy–diagnosis module to provide clinicians with even better diagnostic accuracy. In the spirit of crowdsourcing, Almeida adds that the success of ImageJS will rely upon the willingness of its users to develop their own modules to solve their own specific problems using the game-like interface.

“Funding agencies are also interested in these out-of-the-box solutions. However, there is also some skepticism that will probably need more time to be overcome through clear success stories with these new gaming based solutions to existing problems,” says Molt’s designer Aydogan Ozcan, an associate professor at UCLA.

Despite the funding limitations — and the formidable challenge of designing a game that people will want to play — gaming is gaining traction and enjoying a favorable reception from the research community.

“On a one-to-one basis, everyone seems to love it. They get the potential and are just waiting to see what is going to happen,” Scripps’ Good says. “But developing these games is an enormous challenge. Our background is in bioinformatics, it’s not in making things that are fun. It’s hard to make a fun game all by itself, let alone make one that will solve a difficult problem.”