3D Flex generator using composite maps

Dear cryosparc community,

A complex I am working on has some flexible regions that can individually be refined to sub 3.5A. I was able to get decent composite maps by stitching the global and local refinement maps in Phenix/Chimera. With 3D variability and 3D Flex jobs, I was able to “see” the modes of conformational variability within my complex at low resolution - basically how different parts move. However, I was never able improve the global map significantly with 3D flex.

For the static map, I am, for now, happy with the stitched composite map, but it would be interesting still to interpret the different conformations I see from 3D variability/3D Flex jobs. Specifically, it would be interesting to see relative distances between regions and what interactions maybe allowed/disallowed in different conformations. However, as the 3D flex train job does not really improve the consensus map in the flexible regions, the volume series I get using 3D flex generator is also not very informative by itself for this. So, I am looking for advice on the best approach here to learn about different conformations while avoiding over-interpreting.

The ideal case would have been to somehow use individual particle alignments from the local and global refinement jobs to get a volume series of different conformations, but I couldn’t find a way to do this. So, I ran a 3D flex train job on the consensus refinement output with Nyquist below 8 A (to avoid overfitting), and then used my composite map as the “high resolution input” in 3D flex generator job. This gave me nice volume series with different (and, ostensibly plausible) conformations.

The question, though, is: is this an acceptable thing to do? I reckon, the answer is likely “no”, but even if so, I would appreciate any suggestions on what I could do/how I could approach this instead.