Developing Combined Simulations and Cryo-EM Methods for a Quantitative Understanding of Conformational Landscapes

Many continuous heterogeneity methods generate landscapes with volumes corresponding to regions
with particles, but it is difficult to obtain interpretable and comparable results. Therefore, we have
developed the Scipion Flexibility Hub and a pipeline based around atomic models fitted to the volumes
generated by continuous heterogeneity methods. This approach makes use of the ProDy software for
protein structure and dynamics analysis, enabling landscapes to be projected onto meaningful axes, such
as inter-residue distances and principal components (PCs) of residue motion. The PCs can be visualized in
the normal mode wizard (NMWiz) in VMD. These axes can be easily used for structures coming from
different datasets, the PDB, or molecular simulations, enabling their comparison in a common space.
We have extensively tested this pipeline on various datasets including the spike glycoprotein from two
SARS-CoV-2 variants, enabling us to draw conclusions about the overall variation within both datasets
together as well as differences between them that may underlie variant effects. We compared these
with published structures, revealing potential effects of other conditions such as stabilizing mutations.