Center for Critical Assessment of Structure Prediction (CASP)

University Of California At Davis, Davis CA

Investigators

Linked publications & trials

Abstract

PROJECT SUMMARY CASP (Critical Assessment of Structure Prediction) is an ongoing community experiment to determine and advance the state of the art in computing the structures of biological macromolecules. Experiments are conducted every two years, the results are discussed at an international conference and published in a journal special issue. Approximately 90 research groups world-wide participate. In the most recent CASP15 (2022), there were 53,000 submissions in nine modeling categories, including over 42,000 tertiary structure models. Recent rounds of CASP have found dramatic progress in the accuracy of calculated protein structures, now in many cases rivalling that of experiment. These advances are the outcome of a long process of methods development, nurtured by CASP. The successful approaches are all deep learning based, and it is clear that this technology is likely to have a major impact in other areas of computational structural biology. Accordingly, CASP is undergoing a major restructuring so as to advance these broader applications. The new deep learning areas are the structures of protein complexes; RNA, DNA, and complexes of these with proteins; protein-organic ligand interactions (relevant to drug design); ensembles of conformations for macromolecules (rather than the old paradigm of one sequence/one structure); and assessing the power of integrated modeling, where the computational methods are used to enhance interpretation of low-resolution experimental data. Because of the method improvements, 100s of millions of computed structures are now publicly available. The quality of these is unknown. To address this problem, in a second major innovation, CASP is introducing a system of prospective assessment of the publicly available structures, based on comparison of the computed structures with corresponding experimental ones, as the latter become available. An important component of CASP is community building and outreach and these will be strengthened by a machine learning conference series, CASP special interest groups, and creation of a new help resource. This proposal is for continued support of the Center for CASP, which provides the infrastructure and support for the experiments. Principal tasks include registration and communication with participants; solicitation, characterization, and management of modeling targets; collection and validation of structure models; and extensive numerical analysis of submissions. These operations are supported by a secure and robust data infrastructure. The Center will further develop evaluation, analysis, and display software, and provide access to models and evaluation results. Other aspects of the CASP experiments are supported by in-kind contributions from many members of the community, including members of the organizing committee, independent assessors of results, special interest group chairs, CASP participants, experimental groups that supply target structures and materials, and collaborators who collect additional experimental data.