[phenixbb] FSC curves

Pavel Afonine pafonine at lbl.gov
Fri Mar 19 15:21:13 PDT 2021

Hi Juliana,

> After running phenix real space refine with different refinement 
> strategies.  My map-model FSC curve has some interesting features:
>  1. It does not starts at 1 as it does the map FSC curve, is that ok? 
>     What could be causing this?
which one is it? Is it FSC(half_map1, half_map2) or 
FSC(experimental_map, model_map) ?

>  1. It has a dip around 5-6Å, this deep can be a bit up or down
>     depending the refinement strategy I use, therefore the FSC model
>     at 0.5 can cut the resolution at 3.9 or 6 Å. What this deep mean? 
>     Am I doing something wrong in the refinement?
How different these models are?

>  1. The mask and unmask curves completely overlap, is that ok?
This suggests to me the map is probably already masked or somehow 

>  1. Also, my refinement shows a high clash score (15-20), if I change
>     the nonbonded_weight parameter from 100 (default) to higher values
>     the clash score decreases but the FSC curves gets lower them the
>     FSC curve crosses the 0.5 cutoff at 6A.
When it comes to clashscore, generally the lower the better but zero is 
not the goal.

Looking at one of screenshots I see you have 2.42% of atoms with 
occupancy less than 1. Do you really see partially occupied atoms in 4A 
resolution map? Unlikely! So.. please check those out!

I can run more diagnostics for you if you share maps and models 
(off-list of course!).


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://phenix-online.org/pipermail/phenixbb/attachments/20210319/51a265e9/attachment-0001.htm>

More information about the phenixbb mailing list