journal menu
research papers
 | STRUCTURAL BIOLOGY |
accessReconstructing biological molecules with help from video gamers
aDepartment of Computer and Information Science, University of Massachusetts Dartmouth, Dartmouth, Massachusetts, USA, bOncode Institute and Division of Biochemistry, Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands, and cDepartment of Chemistry and Biochemistry and The Knoebel Institute for Healthy Aging, University of Denver, Denver, Colorado, USA
*Correspondence e-mail: [email protected], [email protected]
This article is part of the Proceedings of the CCP4 Study Weekend 2025.
Foldit is a citizen science video game in which players tackle a variety of complex biochemistry puzzles. Here, we describe a new series of puzzles in which Foldit players improve the accuracy of models in the public repository of experimental protein structure models, the Protein Data Bank (PDB). Analyzing the results of these puzzles showed that the Foldit players were able to considerably improve the deposited structures. We describe a mechanism by which the efforts of the Foldit players can be fed back into the structural biology scientific record by using Foldit results as improved input for the PDB-REDO databank. These efforts highlight the continued need for the engagement of the lay population in science.
Keywords: Foldit; PDB-REDO; citizen science; model curation.
1. Introduction
Our shared knowledge of experimental biomolecular structures is stored in the Protein Data Bank (PDB; Burley et al., 2019
). The accuracy of the models in the PDB is important for multiple reasons. For example, the recent successes in deep-learning approaches to predicting protein structure (Jumper et al., 2021) rely on an accurate PDB for continued development. Large sets of PDB entries are used to study specific aspects of macromolecular structure such as side-chain conformations (Lovell et al., 2000), Ramachandran plot distributions (Sobolev et al., 2020), hydrogen bonding (van Beusekom, Touw et al., 2018), metal-binding sites (Zheng et al., 2017; Putignano et al., 2018), nucleic acid conformations (de Vries et al., 2021; Schneider et al., 2018) and post-translational modifications (Schofield et al., 2024). Importantly, the PDB serves as a resource for biologists and biochemists to develop hypotheses to test in their everyday experiments. Mistakes in structure models can cause scientists to base their hypotheses on incorrect data, which causes delays in scientific progress.
The PDB contains many solvable errors at different levels of complexity (Joosten et al., 2012). The data used in crystallography and cryo-EM – maps of electron density or electron potential, respectively – have limitations as they are the result of an indirect experiment with experimental error and limited resolution. Peaks in the distribution of electron density or electron potential represent atomic positions; thus, scientists can fit atoms to these maps to discover the atomic structure of the macromolecule in the experiment. This requires a combined approach of visual fitting of these maps as well as computational tools to aid in complying with the known rules of chemistry and physics in how molecules are put together. Limitations include low resolution of the data, parts of the maps where artifacts or missing data can cause blurriness in the maps and assumptions in the computational models used to process the primary experimental data, among other causes of low data quality. As a result, the scientists interpreting these data can make mistakes. High attention to detail and the use of verification tools can help to prevent some mistakes (Read et al., 2011), but the sheer quantity of data makes it likely that human error will still persist into published and deposited structure models. Furthermore, the PDB does not require peer review for deposited models and as a result many entries contain errors, ranging from inconsequential to egregious.
For this reason, the PDB-REDO project was begun in 2006 (Joosten & Vriend, 2007). The mission of PDB-REDO is to perform automated re-refinement and rebuilding of the crystallographic structure models in the PDB to improve the accuracy of PDB entries and remove model errors in the process. This venture has been successful, with the PDB-REDO databank now containing over 180 000 entries, many of which have an improved fit to the experimental data and more probable structural features (van Beusekom, Touw et al., 2018).
However, is there still a need for improvement of structure models beyond the automated re-refinement by PDB-REDO? Yes, the procedure in PDB-REDO looks at many structural aspects, but it is limited to model issues that can be handled robustly with a very low risk of making the model worse. Indeed, a search of the PDB for its lowest quality structure models shows that the improvements by PDB-REDO in these cases can be modest at best. They still appear to be relatively low quality after PDB-REDO refinement. This suggests that new approaches are needed.
One possible approach is to enlist the aid of humans to improve the PDB via the players of the biochemistry video game Foldit. Foldit is a citizen science game in which players work on a variety of complex biochemistry puzzles, collaborating and competing to create the best possible structure model (Cooper et al., 2010). Model quality within the game is judged by the Rosetta force field (Leman et al., 2020), combined with other elements, such as the fit to an underlying map from experimental data (Horowitz et al., 2016). Previous competitions have shown that Foldit players can solve both crystal structures and cryo-EM structures with higher accuracy than scientists or computational algorithms. The Foldit players were especially adept at improving Ramachandran space usage and reducing steric clashes while not reducing the fit to data (Horowitz et al., 2016; Khatib et al., 2019). Therefore, it became an obvious question to ask whether Foldit players could also improve the structures already within the PDB and whether this could improve the performance of PDB-REDO in cases with very low starting quality.
To test this possibility, we created a new series of Foldit puzzles, termed Reconstruction puzzles. In these puzzles, Foldit players were given protein structure models and the underlying density maps and tasked with improving the structures. We then tested whether the puzzle solutions outperformed the underlying PDB models as input data to PDB-REDO. In this report, we discuss the results after the first 58 of these puzzles. We also discuss our newly created workflow that automatically connects new puzzle solutions to the PDB-REDO databank.
2. Methods
2.1. Data-set selection
To test whether existing PDB entries could be improved by Foldit players, we selected 58 entries solved by X-ray crystallography with available experimental data. The structures were chosen to primarily be protein-only and to have poor model and/or data-fit parameters. Most structures were chosen by hand using PDBe quality metrics. Primarily, models with poor Ramachandran plots and many steric clashes were chosen within this data set. The models have resolutions ranging from 1.3 to 3.5 Å, were deposited from 1988 to 2022, and contained between 499 and 10 052 atoms.
2.2. Reconstruction puzzle setup
Reconstruction puzzles were prepared by performing five rounds of refinement, including simulated annealing, in Phenix (Liebschner et al., 2019) using the original PDB entry with all ligands and waters removed. This new model was given to the players with a feature-enhanced map (Afonine et al., 2015) for reconstruction within the Foldit application. The tools available include both local and global minimization, distance constraints and interactive chain movement. The puzzles were made available sequentially, with each staying available to users for seven days.
2.3. Puzzle-solution processing
A set of post-processing steps occurred following the completion of each puzzle.
A Python script was used to orchestrate the clustering of player-generated solution files based on their structural similarity, employing the Cα root-mean-square deviation (CA-RMSD) metric, starting with the top-scoring solution for that Foldit puzzle. The score for a puzzle solution is judged by the Rosetta force field (Leman et al., 2020) and the fit to the underlying map from experimental data (Horowitz et al., 2016). Iteratively, the script varies the threshold for CA-RMSD clustering, starting from 1 Å, searching for the next-highest-scoring solution that is at least 1 Å CA-RMSD away, and decreasing that threshold until the total number of distinct clusters is greater than 100. So far, the Foldit community have generated up to 150 000 solutions for a single Reconstruction puzzle. Therefore, this iterative clustering process is crucial to reduce the results to a concise set of models.
Following clustering, the solutions were clustered and ranked based on their Foldit score, with the top 100 clustered solutions moving to the next step. The Python script then submitted the top 100 clustered solutions by score to the API of the PDB-REDO webserver (Joosten et al., 2014), facilitating the submission of these clustered solutions for further refinement and analysis.
The PDB-REDO procedure used here entailed optimizing the weight between the experimental data (as retrieved from the PDB) and restraints for covalent geometry and atomic B factors to maximize the fit to the experimental data whilst maintaining normal geometry. Combinations of different types of additional restraints were used while refining the models with REFMAC (Kovalevskiy et al., 2018): homology-based hydrogen-bond restraints (van Beusekom, Touw et al., 2018) were applied with eight solution sets, general hydrogen-bond restraints with 18 sets, local noncrystallographic symmetry restraints with 26 sets and jelly-body restraints with 19 sets (Murshudov et al., 2011). All of the model-rebuilding steps in PDB-REDO were excluded to stay close to the original solution.
Upon completion of the PDB-REDO calculations, the finalized structure models and associated metrics were retrieved. The top 10 (by Rfree) redone models by were written into new PDB files and the corresponding refinement statistics were organized into comprehensive data dictionaries and graphs, both of which facilitate further analysis. A specialized picker algorithm (Joosten et al., 2012) then performed the final model selection by first rejecting all candidates that had bond-length or bond-angle r.m.s.Z values greater than 1.0. Models that showed clear signs of overfitting based on the ratio between Rfree and the R factor were also rejected (Joosten et al., 2012). The remaining models were sorted by Rfree and the model with the lowest value was selected as the winning model.
2.4. Testing in PDB-REDO
For each PDB entry in the data set two full PDB-REDO calculations were run (including model rebuilding), one with the original PDB model as input and one with the top puzzle solution as input. The resulting models were analyzed with WHAT_CHECK (Hooft et al., 1996), Tortoize (van Beusekom, Joosten et al., 2018) and MolProbity (Williams et al., 2018).
3. Results and discussion
3.1. Reconstruction puzzle results
In total 58 Foldit puzzles were performed. On average, each puzzle had ∼80 000 submitted solutions. After clustering and scoring, the top cluster solutions were submitted to the PDB-REDO server to see how these performed in reciprocal-space refinement. A winning solution was selected based on the algorithm in Section 2.3.
How did the Foldit players improve on these structures? To get an idea, we asked several Foldit players whose structure models were chosen as the best overall structure. The approaches taken by different players were quite variable (see Supplementary Information S1). In most cases, the strategy involved a combination of hand-fitting (interactively moving atoms in the graphical interface) as well as automated tools and scripts. However, in certain cases the top solutions used only scripts and automated tools, suggesting that considerable improvement in automated fitting of structural maps can be accomplished (Supplementary Information S1.2). The top player, Galaxie, improved ten out of 58 puzzles with the highest score. An analysis of their strategy is given in Supplementary Information S1.3.
3.2. PDB-REDO analysis and implementation
The winning solution of each puzzle was then run through the complete PDB-REDO pipeline for comparative analysis to see whether using the Foldit structure models as a starting point is an improvement over using the structure models from the PDB itself. The distribution of model quality metrics was plotted (Fig. 1). In 43 of the 58 cases, the overall structure quality (based on the overall MolProbity score) was improved by using the Foldit structure as the starting point. In general, the Foldit player structures had considerably improved chemical and physical properties in the vast majority of puzzles by multiple metrics (Fig. 1 and Supplementary Information S2). As seen by the MolProbity clashscore, the atomic clashes of the structure models were especially improved, but not in all cases. In terms of Rfree, there was no clear trend towards improvement or deterioration. Eight cases showed a significant change based on the criteria we previously established (de Vries et al., 2021): four cases improved and four deteriorated.
![[Figure 1]](lie5003fig1thm.gif)
|
Figure 1
Comparison of model quality scores for PDB-REDO output using the original PDB model (x axis) or the Foldit-refined model (y axis) as input. Each point is a separate PDB entry. From top left to bottom right: Rfree (from REFMAC), Ramachandran Z-score (from Tortoize), side-chain rotamer normality Z-score (from Tortoize), hydrogen-bond satisfaction fraction (from WHAT_CHECK), fine packing quality Z-score (from WHAT_CHECK) and clashscore (from MolProbity). Emoticons mark the side of the diagonal that indicates an improvement. Cases with a significant change in Rfree when switching to the Foldit model are highlighted in each plot. |
To avoid adding cases to the PDB-REDO databank where the Foldit puzzle solution performs worse in terms of fit to the experimental data than the original PDB model, a model-selection algorithm had to be created that was sufficiently selective but, for efficiency, did not require running PDB-REDO twice. Many decisions in PDB-REDO are based on comparing refinements in which only one parameter (for example a restraint weight or the B-factor model) was changed (Joosten et al., 2012), and we tested whether such a solution was possible in this case. After the initial calculation of R factors for the PDB model, as is the normal procedure in PDB-REDO, both the original PDB model and the Foldit solution are subjected to 20 cycles of restrained refinement in REFMAC with automatic weighting and isotropic B factors. Noncrystallographic symmetry and twinning are considered if required. Because Foldit puzzles are created only for PDB entries with very poor geometric quality and the Foldit scoring function promotes strong geometric improvement, it is assumed that the geometric quality of Foldit models is generally better than that of the PDB model. The best refinement result could therefore be selected with the picker program (Joosten et al., 2012) that focuses on the fit to the experimental data. The Foldit model was selected as the best input model unless (i) the PDB model gave a free log likelihood (as reported for the REFMAC refinement) that was 6.7 points better than the Foldit model, which corresponds to a `decisive' Bayes factor (Kass & Raftery, 1995), and (ii) Rfree for the PDB model was at least σRfree, i.e. one estimated standard deviation of Rfree, lower (with σRfree equal to Rfree divided by the square root of the number of test-set reflections). This approach leaves the option of accepting a small deterioration in Rfree to obtain (much) better geometric quality.
A new PDB-REDO pipeline was created that, for each PDB-REDO run performed on a PDB entry, queries the Foldit database for the existence of a Reconstruction puzzle model. If this exists, the above test was performed. Of the 58 test cases, the Foldit model was selected 42 times (Supplementary Information S2). With respect to always using the Foldit model, this reduced the number of cases of model deterioration in terms of fit to the experimental data (Rfree) from 29 to 21, with an acceptable loss of geometric quality improvement (Fig. 2). For instance, for the overall MolProbity score improvement, the number of cases that improved was reduced from 43 to 33 and the number of cases that deteriorated was reduced from 13 to 10. Other geometric quality indicators followed the same trend (Supplementary Information S2).
![[Figure 2]](lie5003fig2thm.gif)
|
Figure 2
Comparison of model-quality scores for PDB-REDO output using the original PDB model (x axis) or the model selected by picker in PDB-REDO (y axis) as input. Each point is a separate PDB entry. From top left to bottom right: Rfree (from REFMAC), Ramachandran Z-score (from Tortoize), side-chain rotamer normality Z-score (from Tortoize), hydrogen-bond satisfaction fraction (from WHAT_CHECK), fine packing quality Z-score (from WHAT_CHECK) and clashscore (from MolProbity). Emoticons mark the side of the diagonal that indicates an improvement. Cases with a significant change in Rfree when switching to the Foldit model are highlighted in each plot. |
3.3. Yeast mitochondrial import inner membrane translocase subunit TIM44p
There are cases in our test set where changes to the structure can be observed at many points. PDB entry 2fxt (Josyula et al., 2006), the C-terminal domain of yeast mitochondrial import inner membrane translocase subunit TIM44p, which was solved at 3.2 Å resolution, is such a case. When comparing the original PDB model with the model made with Foldit followed by PDB-REDO, a clear change in secondary structure is observed, notably in the central β-sheet (Fig. 3). Analysis with DSSP (Hekkelman et al., 2025) showed that the PDB model had 48 β-strand/β-bridge residues, whereas the updated model had 56. Overall, 121 out of 192 residues had an α/π-helical or β-strand/β-bridge secondary structure. In the updated model this was 134 residues out of 192.
![[Figure 3]](lie5003fig3thm.gif)
|
Figure 3
Comparison of the original PDB model (left) of yeast mitochondrial import inner membrane translocase subunit TIM44p (PDB entry 2fxt; Josyula et al., 2006 ) with the Foldit model after PDB-REDO (right). Top: in the overall structure a clear change is visible, notably in the distribution of β-strands. The number of β-strand/β-bridge residues increases from 48 to 56. Bottom: rearrangement of a surface loop causes a register shift. At the position marked with a star there was originally an Asp residue, whereas in the updated model there is the fully conserved Gly388 residue. Models are shown with their corresponding 2mFo − DFc map contoured at 1.1σ. This figure was made with CCP4mg (McNicholas et al., 2011). |
In many of the cases though the changes are localized, and often in regions of poor electron density such as loops. However, these regions can be important biologically. In PDB entry 2fxt we observe that Gly388 has changed position by 3.2 Å due to a so-called register shift (Fig. 3). This glycine is fully conserved over 1522 sequences in HSSP (Joosten et al., 2011) and this glycine is also conserved in the more distant human ortholog (UniProt O43615). Moreover, AlphaMissense (Cheng et al., 2023) suggests that mutations to this residue are very likely to be pathogenic (minimal score 0.922), indicating that this a key residue in the protein.
3.4. Availability in the PDB-REDO databank
With the success of the Reconstruction puzzle series, we have created a new pipeline in which data from Foldit puzzles can be automatically analyzed and evaluated by PDB-REDO for inclusion in the PDB-REDO databank. On a weekly basis the Foldit database is queried to see whether any new or updated Reconstruction puzzle results are available. If so, the associated PDB-REDO entries are marked for replacement in the next update cycle.
Whenever a Foldit model is used, this is clearly documented in the PDB-REDO metadata. In these cases, if the Foldit player has agreed to have their Foldit name attached to the entry, the PDB-REDO webpage will give credit to the Foldit player whose (redone) structure model is now available for public download. The use of a Foldit model in PDB-REDO entries can also be queried in the PDB-REDO archive manager at https://pdb-redo.eu/archive/ with the property FIUSED set to true. Independently, the Foldit players also maintain a Fandom wiki about the combined Foldit and PDB-REDO resource at https://foldit.fandom.com/wiki/PDB-REDO_Foldit_results.
4. Conclusions
With the success of the Foldit players in improving experimental structure models, the Reconstruction puzzle series within Foldit continues, and will continue, to provide improvements to known structures. With the success of this test, the Reconstruction puzzle series can be used to investigate many questions within the large knowledgebase of structural biology. At present, these puzzles are only for crystallographic structure models, but with the increase in the number of cryo-EM structure models, the same problems also persist there, and can also use the intervention of Foldit players.
With the focus on new artificial intelligence tools and how they can improve fields including structural biology, cases such as this are here to remind us that the greatest untapped potential for science is that of humanity.
Supporting information
Supplementary Information S1: Player strategies. DOI: https://doi.org/10.1107/S2059798325008149/lie5003sup1.pdf
Supplementary Information S2: PDB-REDO statistics. DOI: https://doi.org/10.1107/S2059798325008149/lie5003sup2.xlsx
Supplementary Information S3: Full list of Foldit Players. DOI: https://doi.org/10.1107/S2059798325008149/lie5003sup3.pdf
Footnotes
‡The full list of players is available as Supplementary Information S3.
Acknowledgements
We would like to thank J. Flatten for his assistance maintaining the Foldit platform and the RHPC facility of the Netherlands Cancer Institute for maintaining the PDB-REDO compute infrastructure.
Conflict of interest
The authors declare that they have no conflicts of interest.
Data availability
The Foldit software is available from https://fold.it. The benchmark results are available as supporting information and the final structure models are available through https://pdb-redo.eu.
Funding information
Funding was provided by NSF 2051305 to SH and NSF 2051282 to FK, Horizon 2020 Research and Innovation Program grant No. 871037 (iNEXT-Discovery) to RPJ and by an institutional grant from the Dutch Cancer Society and from the Dutch Ministry of Health, Welfare and Sport.
References
Afonine, P. V., Moriarty, N. W., Mustyakimov, M., Sobolev, O. V., Terwilliger, T. C., Turk, D., Urzhumtsev, A. & Adams, P. D. (2015). Acta Cryst. D71, 646–666.
Web of Science
CrossRef
IUCr Journals
Google Scholar
Burley, S. K., Berman, H. M., Bhikadiya, C., Bi, C., Chen, L., Costanzo, L. D., Christie, C., Duarte, J. M., Dutta, S., Feng, Z., Ghosh, S., Goodsell, D. S., Green, R. K., Guranovic, V., Guzenko, D., Hudson, B. P., Liang, Y., Lowe, R., Peisach, E., Periskova, I., Randle, C., Rose, A., Sekharan, M., Shao, C., Tao, Y.-P., Valasatava, Y., Voigt, M., Westbrook, J., Young, J., Zardecki, C., Zhuravleva, M., Kurisu, G., Nakamura, H., Kengaku, Y., Cho, H., Sato, J., Kim, J. Y., Ikegawa, Y., Nakagawa, A., Yamashita, R., Kudou, T., Bekker, G.-J., Suzuki, H., Iwata, T., Yokochi, M., Kobayashi, N., Fujiwara, T., Velankar, S., Kleywegt, G. J., Anyango, S., Armstrong, D. R., Berrisford, J. M., Conroy, M. J., Dana, J. M., Deshpande, M., Gane, P., Gáborová, R., Gupta, D., Gutmanas, A., Koča, J., Mak, L., Mir, S., Mukhopadhyay, A., Nadzirin, N., Nair, S., Patwardhan, A., Paysan-Lafosse, T., Pravda, L., Salih, O., Sehnal, D., Varadi, M., Vařeková, R., Markley, J. L., Hoch, J. C., Romero, P. R., Baskaran, K., Maziuk, D., Ulrich, E. L., Wedell, J. R., Yao, H., Livny, M. & Ioannidis, Y. E. (2019). Nucleic Acids Res. 47, D520–D528.
Web of Science
CrossRef
PubMed
Google Scholar
Cheng, J., Novati, G., Pan, J., Bycroft, C., Žemgulytė, A., Applebaum, T., Pritzel, A., Wong, L. H., Zielinski, M., Sargeant, T., Schneider, R. G., Senior, A. W., Jumper, J., Hassabis, D., Kohli, P. & Avsec, Ž. (2023). Science, 381, eadg7492.
Web of Science
CrossRef
PubMed
Google Scholar
Cooper, S., Khatib, F., Treuille, A., Barbero, J., Lee, J., Beenen, M., Leaver-Fay, A., Baker, D., Popović, Z. & Foldit Players (2010). Nature, 466, 756–760.
Web of Science
CrossRef
CAS
PubMed
Google Scholar
de Vries, I., Kwakman, T., Lu, X.-J., Hekkelman, M. L., Deshpande, M., Velankar, S., Perrakis, A. & Joosten, R. P. (2021). Acta Cryst. D77, 1127–1141.
Web of Science
CrossRef
IUCr Journals
Google Scholar
Hekkelman, M. L., Salmoral, D. Á., Perrakis, A. & Joosten, R. P. (2025). Protein Sci. 34, e70208.
Web of Science
CrossRef
PubMed
Google Scholar
Hooft, R. W. W., Vriend, G., Sander, C. & Abola, E. E. (1996). Nature, 381, 272–272.
CrossRef
CAS
PubMed
Web of Science
Google Scholar
Horowitz, S., Koepnick, B., Martin, R., Tymieniecki, A., Winburn, A. A., Cooper, S., Flatten, J., Rogawski, D. S., Koropatkin, N. M., Hailu, T. T., Jain, N., Koldewey, P., Ahlstrom, L. S., Chapman, M. R., Sikkema, A. P., Skiba, M. A., Maloney, F. P., Beinlich, F. R. M., Caglar, A., Coral, A., Jensen, A. E., Lubow, A., Boitano, A., Lisle, A. E., Maxwell, A. T., Failer, B., Kaszubowski, B., Hrytsiv, B., Vincenzo, B., de Melo Cruz, B. R., McManus, B. J., Kestemont, B., Vardeman, C., Comisky, C., Neilson, C., Landers, C. R., Ince, C., Buske, D. J., Totonjian, D., Copeland, D. M., Murray, D., Jagieła, D., Janz, D., Wheeler, D. C., Cali, E., Croze, E., Rezae, F., Martin, F. O., Beecher, G., de Jong, G. A., Ykman, G., Feldmann, H., Chan, H. P. P., Kovanecz, I., Vasilchenko, I., Connellan, J. C., Borman, J. L., Norrgard, J., Kanfer, J., Canfield, J. M., Slone, J. D., Oh, J., Mitchell, J., Bishop, J., Kroeger, J. D., Schinkler, J., McLaughlin, J., Brownlee, J. M., Bell, J., Fellbaum, K. W., Harper, K., Abbey, K. J., Isaksson, L. E., Wei, L., Cummins, L. N., Miller, L. A., Bain, L., Carpenter, L., Desnouck, M., Sharma, M. G., Belcastro, M., Szew, M., Britton, M., Gaebel, M., Power, M., Cassidy, M., Pfützenreuter, M., Minett, M., Wesselingh, M., Yi, M., Cameron, N. H. T., Bolibruch, N. I., Benevides, N., Kerr, N. K., Barlow, N., Crevits, N. K., Dunn, P., Roque, P. S. S. B. N., Riber, P., Pikkanen, P., Shehzad, R., Viosca, R., Fraser, R. J., Leduc, R., Madala, R., Shnider, S., de Boisblanc, S., Butkovich, S., Bliven, S., Hettler, S., Telehany, S., Schwegmann, S. A., Parkes, S., Kleinfelter, S. C., Holst, S. M., van der Laan, T. J. A., Bausewein, T., Simon, V., Pulley, W., Hull, W., Kim, A. Y., Lawton, A., Ruesch, A., Sundar, A., Lawrence, A., Afrin, A., Maheshwer, B., Turfe, B., Huebner, C., Killeen, C. E., Antebi-Lerrman, D., Luan, D., Wolfe, D., Pham, D., Michewicz, E., Hull, E., Pardington, E., Galal, G. O., Sun, G., Chen, G., Anderson, H. E., Chang, J., Hewlett, J. T., Sterbenz, J., Lim, J., Morof, J., Lee, J., Inn, J. S., Hahm, K., Roth, K., Nair, K., Markin, K., Schramm, K., Eid, K. T., Gam, K., Murphy, L., Yuan, L., Kana, L., Daboul, L., Shammas, M. K., Chason, M., Sinan, M., Tooley, N. A., Korakavi, N., Comer, P., Magur, P., Savliwala, Q., Davison, R. M., Sankaran, R. R., Lee, S., Tamkus, S., Chen, S., Harvey, S., Hwang, S. Y., Vatsia, S., Withrow, S., Luther, T. K., Manett, T., Johnson, T. J., Brash, T. R., Kuhlman, W., Park, Y., Popović, Z., Baker, D., Khatib, F. & Bardwell, J. C. A. (2016). Nat. Commun. 7, 12549.
Web of Science
CrossRef
PubMed
Google Scholar
Joosten, R. P., Joosten, K., Murshudov, G. N. & Perrakis, A. (2012). Acta Cryst. D68, 484–496.
Web of Science
CrossRef
CAS
IUCr Journals
Google Scholar
Joosten, R. P., Long, F., Murshudov, G. N. & Perrakis, A. (2014). IUCrJ, 1, 213–220.
Web of Science
CrossRef
CAS
PubMed
IUCr Journals
Google Scholar
Joosten, R. P., te Beek, T. A. H., Krieger, E., Hekkelman, M. L., Hooft, R. W. W., Schneider, R., Sander, C. & Vriend, G. (2011). Nucleic Acids Res. 39, D411–D419.
Web of Science
CrossRef
CAS
PubMed
Google Scholar
Joosten, R. P. & Vriend, G. (2007). Science, 317, 195–196.
Web of Science
CrossRef
PubMed
CAS
Google Scholar
Josyula, R., Jin, Z., Fu, Z. & Sha, B. (2006). J. Mol. Biol. 359, 798–804.
Web of Science
CrossRef
PubMed
CAS
Google Scholar
Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., Tunyasuvunakool, K., Bates, R., Žídek, A., Potapenko, A., Bridgland, A., Meyer, C., Kohl, S. A. A., Ballard, A. J., Cowie, A., Romera-Paredes, B., Nikolov, S., Jain, R., Adler, J., Back, T., Petersen, S., Reiman, D., Clancy, E., Zielinski, M., Steinegger, M., Pacholska, M., Berghammer, T., Bodenstein, S., Silver, D., Vinyals, O., Senior, A. W., Kavukcuoglu, K., Kohli, P. & Hassabis, D. (2021). Nature, 596, 583–589.
Web of Science
CrossRef
CAS
PubMed
Google Scholar
Kass, R. E. & Raftery, A. E. (1995). J. Am. Stat. Assoc. 90, 773–795.
CrossRef
Web of Science
Google Scholar
Khatib, F., Desfosses, A., Foldit Players, Koepnick, B., Flatten, J., Popović, Z., Baker, D., Cooper, S., Gutsche, I. & Horowitz, S. (2019). PLoS Biol. 17, e3000472.
Web of Science
CrossRef
PubMed
Google Scholar
Kovalevskiy, O., Nicholls, R. A., Long, F., Carlon, A. & Murshudov, G. N. (2018). Acta Cryst. D74, 215–227.
Web of Science
CrossRef
IUCr Journals
Google Scholar
Leman, J. K., Weitzner, B. D., Lewis, S. M., Adolf-Bryfogle, J., Alam, N., Alford, R. F., Aprahamian, M., Baker, D., Barlow, K. A., Barth, P., Basanta, B., Bender, B. J., Blacklock, K., Bonet, J., Boyken, S. E., Bradley, P., Bystroff, C., Conway, P., Cooper, S., Correia, B. E., Coventry, B., Das, R., De Jong, R. M., DiMaio, F., Dsilva, L., Dunbrack, R., Ford, A. S., Frenz, B., Fu, D. Y., Geniesse, C., Goldschmidt, L., Gowthaman, R., Gray, J. J., Gront, D., Guffy, S., Horowitz, S., Huang, P.-S., Huber, T., Jacobs, T. M., Jeliazkov, J. R., Johnson, D. K., Kappel, K., Karanicolas, J., Khakzad, H., Khar, K. R., Khare, S. D., Khatib, F., Khramushin, A., King, I. C., Kleffner, R., Koepnick, B., Kortemme, T., Kuenze, G., Kuhlman, B., Kuroda, D., Labonte, J. W., Lai, J. K., Lapidoth, G., Leaver-Fay, A., Lindert, S., Linsky, T., London, N., Lubin, J. H., Lyskov, S., Maguire, J., Malmström, L., Marcos, E., Marcu, O., Marze, N. A., Meiler, J., Moretti, R., Mulligan, V. K., Nerli, S., Norn, C., Ó'Conchúir, S., Ollikainen, N., Ovchinnikov, S., Pacella, M. S., Pan, X., Park, H., Pavlovicz, R. E., Pethe, M., Pierce, B. G., Pilla, K. B., Raveh, B., Renfrew, P. D., Burman, S. S. R., Rubenstein, A., Sauer, M. F., Scheck, A., Schief, W., Schueler-Furman, O., Sedan, Y., Sevy, A. M., Sgourakis, N. G., Shi, L., Siegel, J. B., Silva, D.-A., Smith, S., Song, Y., Stein, A., Szegedy, M., Teets, F. D., Thyme, S. B., Wang, R. Y.-R., Watkins, A., Zimmerman, L. & Bonneau, R. (2020). Nat. Methods, 17, 665–680.
Web of Science
CrossRef
CAS
PubMed
Google Scholar
Liebschner, D., Afonine, P. V., Baker, M. L., Bunkóczi, G., Chen, V. B., Croll, T. I., Hintze, B., Hung, L.-W., Jain, S., McCoy, A. J., Moriarty, N. W., Oeffner, R. D., Poon, B. K., Prisant, M. G., Read, R. J., Richardson, J. S., Richardson, D. C., Sammito, M. D., Sobolev, O. V., Stockwell, D. H., Terwilliger, T. C., Urzhumtsev, A. G., Videau, L. L., Williams, C. J. & Adams, P. D. (2019). Acta Cryst. D75, 861–877.
Web of Science
CrossRef
IUCr Journals
Google Scholar
Lovell, S. C., Word, J. M., Richardson, J. S. & Richardson, D. C. (2000). Proteins, 40, 389–408.
Web of Science
CrossRef
PubMed
CAS
Google Scholar
McNicholas, S., Potterton, E., Wilson, K. S. & Noble, M. E. M. (2011). Acta Cryst. D67, 386–394.
Web of Science
CrossRef
CAS
IUCr Journals
Google Scholar
Murshudov, G. N., Skubák, P., Lebedev, A. A., Pannu, N. S., Steiner, R. A., Nicholls, R. A., Winn, M. D., Long, F. & Vagin, A. A. (2011). Acta Cryst. D67, 355–367.
Web of Science
CrossRef
CAS
IUCr Journals
Google Scholar
Putignano, V., Rosato, A., Banci, L. & Andreini, C. (2018). Nucleic Acids Res. 46, D459–D464.
Web of Science
CrossRef
CAS
PubMed
Google Scholar
Read, R. J., Adams, P. D., Arendall, W. B., Brunger, A. T., Emsley, P., Joosten, R. P., Kleywegt, G. J., Krissinel, E. B., Lütteke, T., Otwinowski, Z., Perrakis, A., Richardson, J. S., Sheffler, W. H., Smith, J. L., Tickle, I. J., Vriend, G. & Zwart, P. H. (2011). Structure, 19, 1395–1412.
Web of Science
CrossRef
CAS
PubMed
Google Scholar
Schneider, B., Božíková, P., Nečasová, I., Čech, P., Svozil, D. & Černý, J. (2018). Acta Cryst. D74, 52–64.
Web of Science
CrossRef
IUCr Journals
Google Scholar
Schofield, L. C., Dialpuri, J. S., Murshudov, G. N. & Agirre, J. (2024). Acta Cryst. D80, 647–660.
Web of Science
CrossRef
IUCr Journals
Google Scholar
Sobolev, O. V., Afonine, P. V., Moriarty, N. W., Hekkelman, M. L., Joosten, R. P., Perrakis, A. & Adams, P. D. (2020). Structure, 28, 1249–1258.
Web of Science
CrossRef
CAS
PubMed
Google Scholar
van Beusekom, B., Joosten, K., Hekkelman, M. L., Joosten, R. P. & Perrakis, A. (2018). IUCrJ, 5, 585–594.
Web of Science
CrossRef
CAS
PubMed
IUCr Journals
Google Scholar
van Beusekom, B., Touw, W. G., Tatineni, M., Somani, S., Rajagopal, G., Luo, J., Gilliland, G. L., Perrakis, A. & Joosten, R. P. (2018). Protein Sci. 27, 798–808.
Web of Science
CrossRef
CAS
PubMed
Google Scholar
Williams, C. J., Headd, J. J., Moriarty, N. W., Prisant, M. G., Videau, L. L., Deis, L. N., Verma, V., Keedy, D. A., Hintze, B. J., Chen, V. B., Jain, S., Lewis, S. M., Arendall, W. B. III, Snoeyink, J., Adams, P. D., Lovell, S. C., Richardson, J. S. & Richardson, D. C. (2018). Protein Sci. 27, 293–315.
Web of Science
CrossRef
CAS
PubMed
Google Scholar
Zheng, H., Cooper, D. R., Porebski, P. J., Shabalin, I. G., Handing, K. B. & Minor, W. (2017). Acta Cryst. D73, 223–233.
Web of Science
CrossRef
IUCr Journals
Google Scholar
This is an open-access article distributed under the terms of the Creative Commons Attribution (CC-BY) Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original authors and source are cited.
| STRUCTURAL BIOLOGY |
access
