Analysis for localising the human ndc80-microtubule interface by 'subboxing'

Improving the reconstruction at the ndc80-microtubule interface A refinement of decorated microtubules yielded a reconstruction centered on the microtubule lumen in which the microtubule was oriented along the z-axis and both the ndc80-microtubule interface and dam rings were visible at intermediate resolution. We leveraged the multiplicity of the ndc80-microtubule interface in our data to improve the reconstruction in this region. The positions of visible ndc80-microtubule interfaces in the consensus reconstruction were annotated in napari. A consistent local coordinate system was defined programmatically for each ndc80-microtubule interface in which the new z-axis was perpendicular to the microtubule and the new y-axis was oriented along the protofilament in the +z direction of the consensus reconstruction. A rigid body 3D transformation, composed of a shift and a rotation, was calculated from these data and used to recenter and reorient particles from the consensus reconstruction on each ndc80-microtubule interface was then calculated and then applied to the consensus particle set. Subsequent 3D reconstruction of particles extracted at these positions with their associated orientations yielded a reconstruction centered on the ndc80-microtubule interface in the expected orientation, confirming the correct transformation had been applied and providing an initial reference for subsequent refinement. Particle poses were refined locally in a masked region around the ndc80-microtubule interface in RELION 4.0 to yield a final reconstruction in which this interface was significantly improved relative to the consensus reconstruction. This analysis was implemented in Python using the packages starfile, pandas, NumPy, scipy and einops