Foldit Citizen Science Improves Low-Quality PDB Models Integrated via PDB-REDO

What is it about?

Our knowledge of medicine and biochemistry is largely informed by our knowledge of the intricate 3D shapes proteins take on that allow them to function. However, out knowledge-base of these protein shapes has many errors in it that can cause mistakes in later scientific studies. This work describes how citizen scientists playing the biochemistry video game Foldit are correcting these errors to improve our public knowledgebase of protein structure.

Why is it important?

This research investigates the integration of citizen science contributions, specifically from Foldit game players, into the improvement of crystallographic models in the Protein Data Bank (PDB) and their subsequent impact on the PDB-REDO databank. Ensuring the accuracy and reliability of structural models in the PDB is crucial for downstream applications in biology, biochemistry, and computational modeling, and this study explores new collaborative methods to enhance model quality where automated pipelines reach their limits. Key Takeaways: 1. The study demonstrates that Foldit players, through a combination of manual intervention and automated tools, substantially improved the geometric and physical quality of low-quality crystallographic structure models, as measured by metrics such as the MolProbity score and clashscore. 2. Findings reveal that incorporating Foldit-derived models as starting points in the PDB-REDO refinement pipeline led to improved overall structure quality in the majority of tested cases (43 out of 58), especially regarding chemical and physical properties, though not always in terms of fit to experimental data (Rfᵣₑₑ). 3. The research establishes a selective model integration approach, ensuring that only Foldit solutions providing superior geometric quality without significant loss in data fit are incorporated into the PDB-REDO databank, thereby balancing geometric improvement with experimental data fidelity.