Refinement or Post-Process Maps

[mjaremko]

Hi Kiarash,

Can you provide a suggestion on which volume to use for ModelAngelo? My results vary when using maps from a refinement job (I’m using CS), a post phenix Resolve CryoEM job, and a post DeepEMhancer job. In general, I see the best results with a map from a refinement job. Do you see the same? If so, is this due to the datasets you train your neural network on?

Thanks, Matt

[jamaliki]

Hi Matt,

We observe the best results after automated global sharpening with Relion. I would imagine the same to be true for CS. Generally, over-sharpening seems to hurt us more than under-sharpening. DeepEMhancer, in my hands, seems to make things more difficult for ModelAngelo.

This is all related to how we trained ModelAngelo, which is with postprocessed maps in the EMDB.

Best, Kiarash.