Â鶹ÒùÔº

In a groundbreaking new study, researchers led by the University of Massachusetts Dartmouth have shown that players of the can successfully refine and improve previously solved protein structures—enhancing the accuracy of data housed in the Protein Data Bank (PDB), one of the world's most important scientific resources. The paper is in the journal Acta Crystallographica Section D Structural Biology.

Foldit is a global research collaboration that turns protein modeling into an interactive puzzle game, allowing players from around the world to contribute directly to cutting-edge biochemistry research. Earlier Foldit studies demonstrated that players could outperform both expert scientists and automated computational methods in predicting protein structures.

This latest paper, led by Associate Professor Firas Khatib (Computer and Information Science), analyzes how players improved existing models through a series of Foldit Reconstruction puzzles—tasks based on real protein structures already solved by X-ray crystallography or cryo-electron microscopy and stored in the PDB.

"Science is built on shared discovery," said Khatib, who has been part of the Foldit project since its inception in 2008. "Scientists across disciplines use the Protein Data Bank to guide experiments and develop new technologies. Even small errors in its models can lead to incorrect assumptions down the road. What's exciting is that Foldit players have shown they can proactively improve these structures—strengthening the foundation for future research."

The study found that top Foldit solutions were substantial improvements over the original protein models. In most cases, players enhanced the geometry and reduced steric clashes, and in some instances, improved the fit to experimental data.

The study's supplemental materials also include insights from top Foldit players, who describe their diverse strategies for tackling these puzzles. Notably, two players achieved better results than the original expert models using fully automated methods—scripts, or "recipes," written within Foldit that required no manual human input.

"What's particularly impressive," added Khatib, "is that these automated player-created algorithms improved upon professional results. It shows the creativity and innovation that emerge when you invite the public to participate in real science."