[phenixbb] Can we untoggle molecular weight mean deviation check on ensembles in autoMR?

[Randy J. Read]

By popular demand we (by which I mean Airlie McCoy) have put in a new 
option to bypass the molecular weight deviation check. So in a 
script-driven Phaser run you could add a command like ENSEMBLE ensemble1 
DISABLE CHECK ON. However, there isn't yet an option to disable this in the 
GUI.

That might turn out to be a temporary solution. The reason for the 
deviation check is that an ensemble is assumed to represent a set of 
alternative models for the same component, and if the individual models 
cover significantly different parts of the structure, that assumption is 
violated. Initially we allowed 10% deviation then increased that to 20% -- 
it's a relatively arbitrary number chosen conservatively to be on the small 
side but nonetheless sufficiently forgiving for the test cases we had at 
hand.

When we get a chance, we'll have a look at the consequences of violating 
the assumption that the alternative models cover the same part of the 
structure and do what seems appropriate based on that -- set an 
evidence-based threshold, do away with it altogether or come up with an 
improved way of dealing with ensembles.

Regards,

Randy Read

On Sep 2 2010, James Thompson wrote:

> A postdoc in my lab with many more files inside his Phenix build 
> directory just showed me that the Mwt mean deviation check on input 
> ensembles is hard-coded in Phaser. A seemingly arbitrary value of 20% is 
> set within the file emsembles.cc. So, the answer to my previous question 
> is presently NO. We users cannot untoggle or alter this mean deviation 
> value without some recompile of Phaser. I will play with my Sculpt'ed 
> output from the Mustang server to reduce the Mwt deviations within the 
> ensemble of coordinates that I want to feed to Phaser-MP to get by.
>
>James R. Thompson <thompson.james at mayo.edu>
>
>> Can folks untoggle the molecular weight mean deviation check in
>> phenix.autoMR on the ensembles input?
>> 
>> I want to use an ensemble of coordinates generated by the Mustangs
>> server from a SSM alignment.  Therefore, by design the 10 PDBs input to
>> phenix.autoMR vary greatly in molecular weight from their mean value,
>> but the whole is likely an excellent probe for molecular replacement.
>> However, I am stymied by phenix.autoMR:
>> 
>>  
>>  
>> ------------------------------------------------------------------------------------- 
>> FATAL RUNTIME ERROR: Molecular weight of 
>> /home/jrt04/Desktop/GB1STAS/MUSTANGServer_10_1.6_0_sculpt-coot-0.pdb 
>> (6610.56) deviates more than 20% from the mean (3977.08). This pdb file 
>> may contain domains or water not present in the other files. 
>> -------------------------------------------------------------------------------------
>> 
>> I understand why this check might be of utility to the unobservant.  How
>> can I disable this test?
>> 
>> Many thanks,
>> Jim
>> 
>> -- 
>> James R. Thompson <thompson.james at mayo.edu>
>
>