[phenixbb] Real space refine of long side chains
sazanov at ist.ac.at
Wed Nov 27 05:05:05 PST 2019
As you probably know, due to specific radiation damage cryo-EM maps have
a peculiarity that for most carboxylic side-chains density is absent
(unless those Glu/Asp are involved in salt bridges).
For these reasons the ADP refinement I think may need to be somewhat
differently restrained than for X-ray.
We found that individual (not group) B-factor refinement with
sphere_radius=2.5 and distance_power = 2.3 seems to work well to allow
residues with no side chain density and good backbone density to acquire
high B-factors for side-chains but still reasonable B-factors for backbone.
For that to work we have to convert .ccp4 map to .mtz and use
phenix.refine (just for B-factor refinement only).
If that is run first, then subsequent run of RSR with .ccp4 map does not
try to move those side-chains into backbone anymore.
We find that this is useful to do with maps at least in the 3.0 to 4.0 A
resolution range, not only at low resolution like 4 A. At 2.0 it may be
less useful, not sure.
In usual RSR run B-factor refinement is done only at the end, which may
be suboptimal and indeed it would be better if something similar to the
above was an option in the normal RSR run.
I can send you some example off-list.
Prof. Leonid Sazanov FRS
Am Campus 1
Phone: +43 2243 9000 3026
E-mail: sazanov at ist.ac.at
On 26.11.19 20:02, Pavel Afonine wrote:
> does anyone has a clear example that demonstrates this issue and that
> you can share with me? Having such an example would be most helpful as
> it will allow me to investigate the problem and hopefully address it
> I suspect this can potentially happen at low resolution (perhaps
> 4-4.5A and lower) when main-chain density looks like tubes and there
> isn't much density for side chains, so the main chain density attracts
> side chains. However, with default setting this should not happen as
> there are rotamer restraints. So I'm still not sure what's happening
> and this is why I'm looking for an example.
> On 11/26/19 03:55, Leonid Sazanov wrote:
>> We have addressed exactly this problem by writing a script, in which
>> B-factors are refined first and only then real-space refinement happens.
>> This allows disordered side-chains to acquire high B-factors and so
>> they are not pulled into backbone anymore.
>> This cycle is repeated twice to achieve convergence.
>> Parameters for B-factor refinement (ADP restrains) were optimized to
>> account for consistent lack of density for acidic side chains in cryoEM.
>> It is described in: https://www.ncbi.nlm.nih.gov/pubmed/31492636
>> If needed, I can send a script.
More information about the phenixbb