I had a similar experience (with the validatin report) and concluded it was due to phenix.refine using truncate-extended data but validation using the original intensities without truncate, hence losing all the negative reflections (scary to think a significant portion of my data was originally negative!). However when the annotated pdb was prepared, it used my reported completeness not the lower value from validation. I deposited the output reflections.cif from phenix.refine, and since I started refinement with I's in a .sca file, the final mmtz includes those I's (which I think is good). But then validation uses the I's without truncation and gets lower completeness. Interestingly they still repeated my R values pretty accurately. I guess this would not happen if you first convert I's to F's with truncate and start refinement with the F's, as validation would be forced to use those F's. On 11/29/2018 03:16 PM, Antoine Koehl wrote:
Hi,
I’m in the process of uploading data to the PDB. While AIMLESS had estimated my completeness at 94% (similar to the XTriage value of 93.2%), the internal PDB validation through EDS estimates my completeness to be closer to 83.5%. This is a rather large difference.
Both are using the exact same resolution range (49.57-4.00). Is there something that EDS does differently for estimating completeness?
Please advise,
Antoine _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb Unsubscribe: [email protected]