Development of integrative structure modeling

The National Institute of General Medical Sciences awarded The Regents of the University of California, San Francisco $2.255 million under the Biomedical Research and Research Training program (CFDA 93.859) on September 1, 2026.

The award funds development, application, and dissemination of integrative structure modeling methods for computing structural and spatiotemporal models of biomolecular complexes. The recipient will expand the scope of the open-source Integrative Modeling Platform (IMP) from static protein complex structures to dynamic biomolecular neighborhoods—focal biomolecular systems surrounded by neighboring molecules that diffuse and interact in stable or transient fashion. Work will introduce new model types, scoring functions, model sampling schemes, and assessment methods that incorporate increasingly available experimental data including cryo-electron tomography maps, in situ chemical cross-links, proximity labeling data, and AI/ML protein structure models. By combining information from multiple sources—varied experimental data, physical principles, statistical preferences, and prior models—integrative methods will maximize accuracy, precision, and completeness of output models, particularly for biomolecular complexes difficult to characterize using traditional approaches.

Place of performance is San Francisco, California. The period of performance runs from September 1, 2026 through August 31, 2030.

Abstract

Our broad goal is to develop, apply, and disseminate methods for computing structural and spatiotemporal models of biomolecular complexes. One successful approach, integrative structure modeling, casts the build- ING of such models as a structural sampling problem where knowledge about the assembly is encoded into the scoring function used to evaluate candidate models. We have been developing the opensource integrative modeling platform (IMP) program that provides programmatic support for such modeling workflows. Here, we propose to expand the scope of integrative modeling from static structures of protein complexes to their DY- namic biomolecular neighborhoods. A biomolecular neighborhood in the cell consists of the "focal" biomolecu- LAR system of primary interest, surrounded by other, neighboring biomolecules; in a steady state, these neigh- boring molecules are diffusing and may be interacting with a "focal" system, either in a stable or transient fash- ION. Accurate, precise, and complete modeling of the molecular neighborhood of the focal system can be PAR- ticularly informative about its cellular function. We will introduce new types of models, scoring functions for ranking alternative models based on varied input information, model sampling schemes, and model assess- ment methods. These developments will consider increasingly available experimental data about biomolecular neighborhoods, such as cryo-electron tomography maps, in situ chemical cross-links, and proximity labeling data, as well as AI/ML models of protein structures and molecular networks. By combining information from multiple sources, including varied experimental data, physical principles, statistical preferences, and prior models, integrative methods maximize the accuracy, precision, and completeness of output models. Thus, in- tegrative methods are often applicable to biomolecular complexes that are difficult to characterize using single methods alone. In other words, integrative modeling may facilitate overcoming the sparseness, noise, and am- biguity of any single dataset, in turn allowing contributions by methods that could not currently be otherwise used for structure determination. As a test bed system, we will use the biomolecular neighborhood of the NU- clear pore complex embedded in the nuclear envelope. We will also maximize our impact by disseminating and supporting IMP for use by others. Finally, our efforts will continue to shape the worldwide protein data bank effort on representing, validating, archiving, and disseminating integrative structure models.