![](https://secure.gravatar.com/avatar/4cc2da681ff2333988b7a8ad0df03b55.jpg?s=120&d=mm&r=g)
Hi Joe, Thanks for answering Pavel's question.I think there is no problem in using phenix if you want (whatever works for you). Shya
I think the conclusion may be that you should first try to get good convergence with NCS, and then include TLS. The choice of CNS+REFMAC versus PHENIX is probably not important.
Joe
Pavel Afonine wrote:
Shya,
as PHENIX developer I would be delighted to know what exactly did not work in phenix.refine so you decided to create some extra work for you and follow the detour you mentioned below.
Thanks! Pavel.
On 9/25/09 1:03 PM, [email protected] wrote:
Hi Joe, Try doing NCS refinement in CNS and then when you have done considerable refinement use the final pdb file (when you think your model is complete) for TLS refinement in REFMAC. This has worked for me.You just need to define a TLS unit and use that as an input. Shya
I am using NCS restraints with TLS. It seems that the NCS B-factor restraints are applied to the overall B-factor, rather than just the non-TLS part. Is this correct?
In my structure, the disorder varies quite a bit among equivalent molecules, and I don't get good results unless the b_factor_weight is about 1e-3 or less.
I am experimenting with NCS restraints in CNS, where each pair of coordinates allows an LSQ fit of B-factor scale and offset to be excluded from the B restraints. It would probably work even better with TLS, which CNS does not have. I don't know if anyone has tried this, but I think that overall molecular B-factor differences are fairly common.
Joe Krahn
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb