Increasing Public Involvement in Structural Biology

Structure Commentary

Seth Cooper,1,* Firas Khatib,2 and David Baker2 1Department of Computer Science 2Department of Biochemistry University of Washington, Seattle, WA 98195, USA *Correspondence: [email protected] http://dx.doi.org/10.1016/j.str.2013.08.009

Public participation in scientiﬁc research can be a powerful supplement to more-traditional approaches. We discuss aspects of the public participation project Foldit that may help others interested in starting their own projects.

It is now easier than ever for the public to We’re very excited about the possibility Openness to Collaboration get involved in science. The Internet has for games and other forms of public in a Variety of Forms made it feasible for research groups to involvement in science to help advance The core of the project has been a very easily connect with people all over the the field. To our knowledge, there have fruitful collaboration between the Com- world. Personal computers have also been a few other projects actively puter Science and Engineering Depart- become powerful enough to run compu- involving the public in structural biology, ment and the Biochemistry Department tationally intensive programs, giving the and we look forward to many more in at the University of Washington. Both public the opportunity to contribute to the future. Structural biology problems departments were able to bring their scientific research. Volunteer computing involving the analysis of existing mole- knowledge and skills together to make a allows the public to share their spare cules and the design of new ones are successful team. We have also been CPU cycles with structural biology re- promising areas for citizen science as lucky to work with scientists who are searchers in projects like Folding@home structural problems both are amenable interested in collaborating with the public (http://folding.stanford.edu) and Roset- to human spatial-reasoning skills and and allowing them to share credit for the ta@home (http://boinc.bakerlab.org). can make enjoyable puzzles. In the spirit discoveries they have made, not fearing Actively involving the public in scientific of helping others who might be interested that their thunder will be stolen by Foldit research—often referred to as citizen in starting their own such projects, here participants. science—provides a means of engaging are some elements that were important We have also found that participants people’s skills, rather than just their in the development of Foldit. may prefer to be credited differently than computational resources. This approach academics—citizen scientists are a lot has been successfully used in astronomy, Single Quantitative Metric less interested in coauthorship. The three for example, to locate celestial objects of Success Foldit participants whose model led to the with the Galaxy Zoo project (Lintott Foldit is based on the principle that solution of the MPMV-PR all declined et al., 2009). But how can we increase proteins fold to their lowest free-energy coauthorship, choosing instead to publish the ability for the public to get actively states. In structure prediction problems, under their Foldit team’s name. The solu- involved in structural biology research? Foldit participants seek to find the tion of the previously unsolved crystal Citizen science projects can require a lot structure with the lowest Rosetta energy structure, arguably our most prominent of work to set up and maintain, but there (Leaver-Fay et al., 2011) for the sequence result, also came out of a collaboration is the potential to yield great results if they are given. Participants collaborate with another biochemistry lab at the more labs start them. and compete to find the lowest free Adam Mickiewicz University in Poland. For the past several years, we have energy (in the case of Foldit, the highest Using the Foldit-generated model, the been running the Foldit project (http:// scoring) structure and, in design prob- lab was able to solve the crystal structure fold.it), which allows participants to lems, the lowest energy sequence. via molecular replacement. directly manipulate proteins in an online Biology problems that can be posed as video game (Figure 1). In that time, partic- global optimization problems, with a sin- Community Support and Fostering ipants of Foldit have made contributions gle metric such as energy, are more a Connection to Science to a number of scientific publications. amenable for gamification than those and the Project They were instrumental in the solution of that are not. None of the results produced by Foldit the crystal structure of the Mason-Pfizer Having software like Rosetta already in would have been possible without the Monkey Virus Retroviral Protease (Khatib place for computing energies, and gener- hard work of its participants. Because et al., 2011) and the design of a novel ating and sampling alternative structures of this, a large effort is put into sup- synthetic enzyme for the Diels-Alder reac- and sequences, gave us a starting point porting and fostering the development tion (Eiben et al., 2012). for building on a gaming interface. of the community. Participants can

communicate with each other allowed us to continually and the team through a vari- refine the game’s utility as ety of channels, including an a problem-solving tool even in-game chat and forums. as the participants them- In addition to the usual selves become better prob- support, we strive to give lem solvers. participants a connection This cycle also means that to the science behind the there is always something game. Blog posts, videos, new going on with the project. and podcasts from scientists Every week there are new help us to communicate puzzles to solve, and we what the participants are frequently post updates to working on and why it’s the game itself with new fea- important. We also have reg- tures, fixes, and responses ular scientist chats and to participant feedback. We developer chats where par- try to keep the project ticipants can communicate refreshing and alive and so directly with the people that when participants come working on the project to get back there is something new a view of what is going and interesting. on behind the scenes and have any burning questions Development of an answered. This effort has Accessible Interface grown over time with the pro- to Complex Structures ject, and we have added and Problems more channels of information Protein structures can be flow between the participants complex to look at. Rather and project team. We’ve than looking at them in a even recently hired a commu- Figure 1. Involving the Public in Science Can Be Both Rewarding for more traditional way, we nity liaison to help manage Researchers and Fun for Participants wanted to take a new look at this and improve the flow of proteins to cast problems in information. a manner that can involve When the project started, we were Giving more experienced participants the public. We sought to abstract away fortunate to be able to draw on an existing the opportunity to share what they’ve many details and highlight areas of the community to bootstrap ours rather learned and help beginners has also structure where action could be taken. than starting entirely from scratch. Some been useful. The participants have Having people try out this new interface of our earliest participants came from come up with many interesting ways as early as possible was of the utmost the Rosetta@home community, a group to play the game, so we also support importance. Early versions of the interface already interested in protein folding. the sharing of their expertise. Initially, the went through many rounds of playtesting. New projects might look to existing participants themselves created a wiki Even having just a few novice users communities of related interest when (http://foldit.wikia.com) to share their try out the interface can illuminate how starting out. folding strategies. We’ve also added the people want to interact with it and ability for participants to encode their highlight the most critical areas for Support for Social Play and Sharing strategies and share them in the game, improvement: where participants would of Expertise and we’ve seen that they use a variety of get stuck, confused, or frustrated. We wanted to apply the power of multiple techniques. We have found that the game’s inter- people’s minds working together to solve face can be useful for protein structure problems. Foldit participants are not just A Cycle of Continuous Refinement modeling in general. Therefore, we working on structures in isolation; they Based on Data have released a version that is free have the option to form groups, in which From the beginning, we didn’t plan on for noncommercial use, called Foldit they can share structures to get ideas getting everything right the first time. Standalone. It can be downloaded from and build upon each others’ work. Our So we set the project up with the the University of Washington Center for best results have come out of multiple infrastructure and mindset that it would Commercialization website (http://tinyurl. participants working together. It would evolve. com/academic-foldit). be interesting to explore other formats Every week, the project team meets In the last few years, other success- for information sharing, and we hope to to look at structural data gathered in the ful games in structural biology have add functionality for multiple participants preceding week, evaluate the data, and emerged. EteRNA (http://eterna.cmu. to work together simultaneously on the feed that back into the plans for what edu) is allowing participants to help same structure. to do next. This iterative process has design a library of new RNA shapes.

Participants in EyeWire (http://eyewire. been impressed by Foldit participants’ Kawrykow, A., Roumanis, G., Kam, A., Kwak, D., org) are helping to map the neurons of creativity and ingenuity. For the right Leung, C., Wu, C., Zarour, E., Sarmenta, L., Blanchette, M., and Waldispu¨ hl, J.; Phylo players. the retina. Phylo (http://phylo.cs.mcgill. types of problems, cast in the right way, (2012). PLoS ONE 7, e31362. ca) participants are working to solve there is the potential to involve a new Khatib, F., DiMaio, F., Cooper, S., Kazmierczyk, the multiple sequence alignment (MSA) group of creative problem solvers. M., Gilski, M., Krzywda, S., Zabranska, H., problem and were recently shown to Conversely, games such as Foldit can Pichova, I., Thompson, J., Popovic, Z., et al.; be able to improve on automated be a powerful means for popularizing Foldit Contenders Group; Foldit Void Crushers Group. (2011). Nat. Struct. Mol. Biol. 18, 1175– algorithms for MSA (Kawrykow et al., and disseminating structural biology 1177. 2012). research to the general public. Leaver-Fay, A., Tyka, M., Lewis, S.M., Lange, O.F., We encourage anyone working on Thompson, J., Jacak, R., Kaufman, K., Renfrew, difﬁcult structural problems to reach out P.D., Smith, C.A., Shefﬂer, W., et al. (2011). 487 to the public for help. Certainly, involving REFERENCES Methods Enzymol. , 545–574. the public is not a silver bullet to solve Lintott, C., Schawinski, K., Keel, W., van Arkel, H., every problem, but public involvement Eiben, C.B., Siegel, J.B., Bale, J.B., Cooper, S., Bennert, N., Edmondson, E., Thomas, D., Smith, Khatib, F., Shen, B.W., Players, F., Stoddard, D., Herbert, P., Jarvis, M., et al. (2009). Monthly can be a powerful means to complement B.L., Popovic, Z., and Baker, D. (2012). Nat. Bio- Notices of the Royal Astronomical Society 399, existing methods. We have consistently technol. 30, 190–192. 129–140.