Collaborative Research: Tools for Tackling Preferred Specimen Orientation in Cryo-EM

The advent of high-resolution cryogenic electron microscopy (cryo-EM) has fundamentally changed biology by making it possible to determine high-resolution structures of proteins that were previously inaccessible, while also allowing researchers to visualize them in multiple functional states. Advances in cryo-EM technologies have therefore accelerated fundamental discoveries in biology and driven structure-based drug discovery to expand our understanding of molecular mechanisms in health and disease. Cryo-EM requires freezing the biological specimen of interest within a thin hydrated layer on specialized grids. However, the application of cryo-EM is often challenging because protein specimens frequently adopt a limited set of orientations within the vitrified ice, resulting in a preferred orientation that may preclude obtaining high-quality high-resolution maps. This problem is pervasive. It increases susceptibility to bias and overfitting and can ultimately prevent accurate structure determination altogether. Although several tools and techniques quantify the effects of preferred orientation, there is an incomplete mathematical understanding of how to characterize and correct for these effects. This proposal will establish both theoretical and experimental foundations that will characterize and correct the effects of preferred orientation, which will be essential for improving map quality, validation, and structural interpretation, and in turn will maximize the impact, utility, and reach of cryo-EM projects nationally and globally. The goals will address multiple NSF Broader Impact pillars, including STEM education, public engagement, competitiveness in the STEM workforce, and novel infrastructure development. A major barrier to routine high-resolution structure determination remains due to preferred specimen orientation, whereby biological specimens adopt only a limited set of views on the cryo-EM grid. The resulting uneven sampling of projection orientations in 3D space leads to directional resolution anisotropy, reduces global resolution, and can introduce reconstruction artifacts and bias during map refinement. Despite its prevalence, the full impact of uneven sampling on cryo-EM reconstructions and refinements remains poorly understood. This proposal aims to combine theoretical and experimental approaches to define how uneven sampling affects cryo-EM reconstructions and to develop computational strategies that mitigate these effects. Specifically, this work aims to determine and reduce the impact of uneven sampling on directional resolution and map isotropy, improve estimates of global resolution, reduce overfitting during iterative refinement of cryo-EM maps, and establish quantitative best practices for map validation. By combining applied mathematics with experimental cryo-EM, this project will establish a rigorous framework for understanding and correcting uneven sampling distributions, yielding more accurate reconstructions, improved structure interpretation, and greater efficiency in cryo-EM workflows, which would help in formulating better biotechnology applications and strategies on protein engineering. The proposal will yield a framework for integrating sampling tools that can be employed on-the-fly during data collection for improved reconstruction of cryo-EM maps. Beyond cryo-EM, the underlying principles may generalize to other imaging and data-analysis disciplines that reconstruct information from noisy, incompletely sampled measurements. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.