[phenixbb] 1.5.2 and validation

[Nathaniel Echols]

On Oct 23, 2009, at 6:47 AM, Mark Collins wrote:
> I just installed 1.5.2 and tried it out for a few hours this  
> afternoon. Is it
> possible to run the phenix programs from the command line without  
> being forced
> to launch the shiney new gui?

All but a few of the programs have command-line versions - the  
documentation probably needs to be updated to name these.  The  
exceptions in 1.5-2 are the reflection file editor and Polygon.   
However, the next version will have a command-line version of the  
reflection file editor.

The integrated validation tools are also GUI-only but all of the  
underlying components have their own commands:
phenix.model_vs_data (which also summarizes geometry statistics from  
the other programs)
phenix.ramalyze
phenix.rotalyze
phenix.cbetadev
phenix.clashscore
phenix.reduce (not a validation tool, but it identifies Asn/Gln/His  
flips)
phenix.show_geometry_outliers
phenix.real_space_correlation

Except for the last, these are very simple to use.  There will be  
command-line versions of the GUI validation tools in the next version,  
but they really just combine the output of existing programs.

> I have 3 structures I had considered finished by R/Rfree and  
> Molprobity. I was
> playing around with polygons (great way to quickly assess the  
> structure) and
> found that 2 of the 3 strucutres are "bad" with very low RMSDs for  
> bonds,
> angles, chirality, dihedrals and planarity.  All were refined with  
> phenix 1.3,
> but they are all fairly small ~80-90 residues. So do I need to go  
> back and
> revisit these, advice much appreciated?

Actually, there's nothing inherently "bad" about very low geometry  
RMSDs, but (at least at moderate-to-high resolution) they can indicate  
that you've restrained the geometry too tightly, which can make it  
more difficult to refine.  If the R-factors, maps, and validation  
criteria look good, I don't think you need to worry.

> Also I found that I can edit the validate_0.eff (from phenix/ 
> project_data/ )
> and run the file from the command line.   But is there a way to  
> actually change
> keys_to_show and fiters for polygon in the gui?

keys_to_show, yes, but you need to use the separate Polygon program  
rather than the main validation GUI, which uses a pre-set list of  
keys.  The filters can't be changed yet.  Again, this will all be  
fixed in the next version.

> Also it seems that the *twinned crashed the validation when run this  
> way (bug
> reported).


I was wondering what that error meant. . .  You need to specify at  
least a maximum or minimum value - although I'm not sure what these  
should be for that filter.  I think I'll change that error message to  
be more informative (it isn't really a "bug", but the GUI doesn't know  
that).

--------------------
Nathaniel Echols
Lawrence Berkeley Lab
510-486-5136
NEchols at lbl.gov