Re: [ccp4bb]: Refmac vs. cns

[Anastassis Perrakis <perrakis@nki.nl>]

***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***

>> I am very interested in this comparison too. Can anybody give me some
>> details about the comparison between Refmac5 and CNS? I have never use
>> Refmac5 before. Can refmac5 do simulated annealing?
>> 

>  Heres a fire storm brewing..
>  Can some-one comment on how valuable simulated annealing is at various
> resolutions?
> 
>  In examples at better than 2.5A we usually seem to finish up doing just
> about the same amount of rebuilding as you do after a ML run. 

Ok, here it goes, lets hope it will get 'lively'
Not directly an answer on simulated annealing but very related.

At resolutions better than ~2.0 there is not doubt in my personal (and 
slightly biased, see also Warren/PyMol) opinion that ARP/wARP with 
Refmac(5) will outperform simulated annealing in CNS (or 
Refmac-on-its-own). It takes (significantly) longer, but no longer than 
a few hours that can be used for coffee/beer/sleep/more-hard-work (use 
the last option carefully, it can be bad for you). Most of the 
rebuilding can be eliminated by doing the autobuilding which will create 
a very good model (nearly perfect really) for the bits that it does 
build - lets sa 80-95% of the structure. You have to do the rest in your 
own.

And as a reminder there IS in ARP/wARP 5.1 a script/program to build and 
real-space refine side chains as well - most people seem to ignore that 
(there are at least two ActaD papers that explicitely state that 
'ARP/wARP does not build side chains' !!!)

I am not 100% convinced of what to say for 2.2 - 3.2. There had 
definately been cases that ARP/wARP with Refmac has produced at that 
resolution range (even at 3.2, but with high solvent content) MAPS that 
were far better than various tries with CNS and simulkated annealing. 
Especailly when starting from incomplete models. HOWEVER, CNS produced 
the BETTER model ! (ARP/wARP will throw away atoms it does not like so 
the model is BAD !) Thus I would suggest to get the map with ARP/wARP - 
REFMAC and the model either from refmac(5) alone (for resolutions not 
much worse than ~2.7) or from CNS (especially if resolution is worse 
than ~2.7)


		Tassos