Hi Lisa
For the validation of my model to my map in phenix 1.17.1 I tried to refine my shaken model (0.5Å) against my half map in phenix real space refine.
first off, I think this way of validation is quite faulty conceptually... How you know 0.5A is the best number and say not 0.7 or 0.3? By shaking the model, you validate the shaken (perturbed) model, which is not the model you are publishing and reporting statistics for. If you shake the model twice with same shake doze, you will get two different models with the same amount of perturbation, which means numbers you derive from such models can (and likely will) be different. So... with this in mind I'm not really seeing any use of such a validation approach; this does not make sense to me at all! Instead, why wouldn't you just follow the standard validation procedure, where you validate the model, the data and the model-to-data fit? The Phenix tool for this is "Comprehensive validation (cryo-EM)", it is available in the GUI and it is one stop to do all of this. And if you want to find uncertainties in atomic positions due to refinement converging to local minima, you can use phenix.mia tool that will generate and refine many perturbed models using exact same way and give you an ensemble of models that can show local variations. All of the above described here: https://www.ncbi.nlm.nih.gov/pubmed/30198894 Good luck and let me know if you have any questions! Pavel
However, whatever I try I get a weird behaviour in the FSC(Model-map), see the image below. It seems not to be able to reach 0 in the correlation, but keeps getting stuck at 0.1.
I tried various things to solve this, but unfortunately nothing helps: - Using the other half map - Cut of the heterogeneous part of my model and map and refine this - Used refined data without using a mask for the last iterations - Used different resolutions to refine against - Tried data from someone else in the lab (same microscope, same imaging conditions), to check if I am doing something wrong (this gave good results)
When doing the same refinement in Refmac5 in ccpem 1.4 it does seem to work, however I would not like to switch programs, because this will mess up my final model.
Best, Lisa
