Nat. Struct. Mol. Biol., published online 18 September 2011, doi: 10.1038/nsmb.2119

Researchers have created several online games in attempts to harness efforts from the general public to advance scientific knowledge, but it is often difficult to quantify exactly what is gained from these efforts. In Foldit (http://fold.it/), players attempt to identify low-energy structures for protein sequences using both computational tools and their own three-dimensional acuity. Now Khatib et al. report the specific success of Foldit players both in a protein-structure prediction competition (CASP9) and in solving the structure of a protein de novo. In CASP9, Foldit players took on challenges in several categories including template-based modeling, template-free modeling and refinement. In template-free modeling, the Foldit Void Crushers group improved a structure generated by the Rosetta Server to generate a near-native conformation of Pseudomonas aeruginosa protein T0581 as determined by comparison to the subsequently released structure. The Foldit community provided a bigger surprise in a second 'real-world' challenge. Although there is crystallographic data available for the Mason-Pfizer monkey virus retroviral protease monomer, molecular replacement and other strategies have been unable to phase the data, frustrating scientists for more than a decade. After Khatib et al. issued a call for help, the Foldit Contenders group returned a model that could be used to solve the crystallographic data, yielding a final refined structure within a few days.