Research Project
10242780
EMAN is one of the most well-established and widely used scientific image processing suites targeting the rapidly growing CryoEM/CryoET community worldwide. In turn, the CryoEM and CryoET studies which it enables permit determination of the structures of interacting macromolecules both in-vitro and in-vivo, and are being used to better understand the biochemical processes taking place in cells, to better identify potential drug targets and develop novel diagnostics. With the higher resolutions now possible in this field, direct drug interaction structural studies are now possible, and being used to gain insight into the mode of action of drugs within the cell. Unlike many newer tools in the field, such as Relion, CisTEM and CryoSparc, which focus on specific refinement tasks, EMAN is a versatile, modular suite capable of performing a variety of image processing tasks with hundreds of algorithms supporting virtually all of the standard file formats and mathematical conventions used in the field, as well as other related imaging fields. It provides an ideal platform for prototyping fundamental new algorithm developments, while still able to achieve data-limited resolution in single particle reconstruction. While high resolution single particle refinement has become routine in recent years, thanks largely to the dramatic data quality improvements provided by new detector technology, there remain significant opportunities for improvements in mitigating model bias, efficient use of data, and analysis of complexes with compositional or conformational variability. Some of the most important problems from a biological perspective involve the sort of compositional and conformational variability which remain challenging problems. The field also remains susceptible to problems of initial model bias, which are exacerbated in systems exhibiting structural variability, and as a result many structures are still published with exaggerated resolution claims. The standard protocols used by many in the field typically involve discarding a very large fraction of the raw data (as much as 80-90% in some cases), often based on qualitative assessments, raising questions related to rigor and reproducibility of structural results. In this proposal, we will develop or adapt image processing techniques to help resolve these issues, based on developments or unrealized concepts from mathematics and computer science.
