This article was accepted into the corpus but its outbound wikilinks were never NER-processed — typical at the deepest BFS hop or when the run's entity cap was reached. No expansion funnel to show.
| CASP | |
|---|---|
| Name | CASP |
| Established | 1994 |
| Frequency | Biennial (historically) |
| Discipline | Structural biology, bioinformatics, computational chemistry |
| Location | International (rotating hosts) |
CASP
CASP is a community-driven benchmarking initiative that evaluates predictive methods for protein structure and related problems. It brings together method developers, experimentalists, and users from institutions such as European Molecular Biology Laboratory, Roskilde University, University of Cambridge, University of Oxford and Massachusetts Institute of Technology to assess advances in computational structure prediction. The experiment influences research at centers including European Bioinformatics Institute, Johns Hopkins University, Stanford University, Harvard University, and Broad Institute by providing blind tests, datasets, and standardized metrics.
CASP operates as a blind, community-wide experiment: organizers solicit unpublished targets from structural biology laboratories like Diamond Light Source, European Synchrotron Radiation Facility, Brookhaven National Laboratory, and Argonne National Laboratory, then invite teams from groups such as DeepMind, University of Toronto, University of California, San Francisco, University of Chicago, Max Planck Institute for Biophysical Chemistry, and Riken to submit predictions. Results are assessed against experimentally determined structures from projects at European Molecular Biology Laboratory — European Bioinformatics Institute, Los Alamos National Laboratory, National Institutes of Health, and specialist facilities including Oak Ridge National Laboratory and Scripps Research. Performance is summarized using metrics developed in concert with contributors from University of California, Berkeley, Columbia University, University of Washington, and University of Illinois at Urbana–Champaign.
CASP was launched in 1994 by researchers associated with University College London, Working Group on Protein Folding, and collaborators at European Molecular Biology Laboratory. Early participants included groups from University of California, San Diego, University of Pennsylvania, University of Cambridge, Weizmann Institute of Science, and Institute Pasteur. Milestones include the adoption of template-based modeling assessments influenced by work at Protein Data Bank, Brookhaven National Laboratory, and the shift to include modeling of domains and assemblies driven by inputs from European Synchrotron Radiation Facility and National Synchrotron Light Source. Over successive rounds, contributions from teams at Rosetta Commons (originating at University of Washington and University of North Carolina at Chapel Hill), Zhang Lab ( University of Michigan connections), and industry entrants like DeepMind reshaped expectations for fold recognition and free modeling.
CASP’s workflow integrates target acquisition, prediction submission, evaluation, and analysis. Target acquisition relies on collaborations with structural groups at European Molecular Biology Laboratory, Institute of Protein Research (Russia), Max Planck Institute for Developmental Biology, and national labs such as Lawrence Berkeley National Laboratory. Prediction submission pipelines include methods developed at University of Tokyo, Seoul National University, Tsinghua University, ETH Zurich, and University of Cambridge. Assessment instruments use metrics such as GDT, TM-score, and LDDT, with methodological contributions from researchers at University of California, Santa Cruz, University of Helsinki, and Université Paris Diderot. Components of the experiment encompass tertiary-structure prediction, model quality assessment, contact prediction, refinement, and complex modeling; each subcategory attracts teams from EMBL-EBI, European Molecular Biology Laboratory, Cold Spring Harbor Laboratory, and MRC Laboratory of Molecular Biology.
CASP has driven practical advances recognized by venues including Nature, Science, Proceedings of the National Academy of Sciences, Cell, and field-specific journals such as Journal of Molecular Biology. Notable outcomes include improvements in homology modeling that benefited projects at Protein Data Bank, enhancement of contact prediction algorithms used by groups at Max Planck Institute for Developmental Biology and Weizmann Institute of Science, and breakthroughs in deep learning applied by teams from DeepMind, University of Toronto, University of Oxford, and Institute for Protein Design. These results influenced structural genomics initiatives at Joint Center for Structural Genomics, therapeutics research at Novartis, Roche, GlaxoSmithKline, and vaccine design efforts at Imperial College London and Scripps Research.
CASP’s community spans academic labs, industry groups, and national facilities. Academic participants include University of California, Davis, University of Alberta, Peking University, Australian National University, and McGill University. Industry contributors come from Google DeepMind, Schrödinger, Inc., Exscientia, and biotech firms collaborating with Biogen and Amgen. Supporting organizations and funders include Wellcome Trust, European Research Council, National Science Foundation, National Institutes of Health, and agencies such as Japan Society for the Promotion of Science. The community convenes at workshops hosted by European Molecular Biology Laboratory, Cold Spring Harbor Laboratory, and conference partners like ISMB, RECOMB, and Gordon Research Conferences.
Critiques of CASP address target selection biases tied to contributor networks at Protein Data Bank and national labs, potential overfitting toward metrics developed with input from University of California, Santa Cruz and University of Washington, and uneven representation of groups from regions such as Africa and parts of South America. Limitations include challenges in assessing membrane proteins studied at Membrane Protein Laboratory (Diamond Light Source), complexes relevant to European Molecular Biology Laboratory collaborations, and dynamic systems probed at Max Planck Institute for Biophysical Chemistry. Debates have involved stakeholders at Nature Methods, Science Advances, and organizers with ties to University College London.
Future CASP iterations plan expanded integration of cryo-EM targets from European Synchrotron Radiation Facility and National Center for Electron Microscopy, enhanced assessment of protein–protein interactions with input from EMBL-EBI and Riken, and broader participation initiatives engaging researchers at African Academy of Sciences, Latin American Structural Biology Network, and Asian Infrastructure Investment Bank-funded centers. Methodological frontiers include combining deep learning advances from DeepMind and University of Toronto with physics-based refinement developed at Rosetta Commons and Max Planck Institute for Biophysical Chemistry.
Category:Protein structure prediction