Hi Kate,

one can endlessly talk about this matter and there is a number of good papers out there.

At some point I spent significant amount of time trying to address the questions you mention below. I selected all structures from PDB for which NCS is available and re-refined them with and without NCS. My goal was to answer this question: given a model and data, can I tell for sure if I need to use NCS or it's better to refine without using NCS. In the end I couldn't derive any reliable formula to be able to decide if I need to use NCS or not. All I re-discovered was the old and well known fact: if have have NCS, then lower the resolution, more likely you want to use NCS -:) Obviously, at ~1.5A resolution the use of NCS was only damaging. And ~2.0A is kind of "gray area".

So this is why I wrote in my previous email that it is much easier, quicker and reliable to just try both ways and see which one works better, rather than waste time on speculating -:)

Cheers,
Pavel.

On 8/25/2008 11:33 AM, Kate Kavanagh wrote:

I was wondering if there a consensus of when it is mandatory to use ncs 
in refinement. Is is dependent on:

1. resolution?
2. data-to-parameter ratio?
3. difference between Rcryst & Rfree?
4. something else...

Or does one just do as Pavel states and try both ways ?

Regards,
Kate



Pavel Afonine wrote:

Hi Francis,

Also if anyone has a paper/protocol for NCS refinement, I'd love to  
check it out (to figure out which map to fit to).

Just something to keep in mind... I always prefer to run two refinements 
in parallel with and without using NCS. It's resolution (and 
data-to-parameters ratio) dependent. I've seen a  number of cases where 
using the NCS is very beneficial and similar number of cases where it's 
better to not use NCS (mostly at resolutions ~2-2.5A).

I would also check manually the NCS groups that are found by 
phenix.refine automatically. I have some cases where automatic detection 
did not yield a good selection.

Pavel.

_______________________________________________
phenixbb mailing list
[email protected]
http://www.phenix-online.org/mailman/listinfo/phenixbb

_______________________________________________
phenixbb mailing list
[email protected]
http://www.phenix-online.org/mailman/listinfo/phenixbb