Hi Dave,
Sorry for the delayed response, I have just returned from holidays.
Please see below:
On 4 June 2015 at 17:02, Lawrence David Finger <d.finger@xxxxxxxxxxxxxxx>
wrote:
I posed these questions some time ago, but I have not heard from anyone or
whether my message was accepted by the moderator. I guess I am not on the
list. Below is what I wrote initially.
I just checked the mailing list backend, and your d.finger email
address is not present. You should try subscribing again. One step
that is often forgotten is that after you receive the subscription
request email and follow the link, you have to in that page
specifically select to subscript or ignore the subscription. In most
cases when you are not subscribed, the mailing list maintainer Chris
MacRaild will see your message and accept it. One problem though is
that public mailing lists are a massive magnet for spam. So if your
message is buried in a pile of 1000 spam messages, it is sometimes not
seen and thrown out with the 1000 junk messages. Being subscribed to
the list avoids this.
We have a question concerning our results from our latest calculations
using relaxGUI. We have fed the program NMR relaxation data at 500 and 600
MHz on the free protein and a structure of the protein in complex with DNA
from an X-ray structure. We know that binding of DNA induces some secondary
structures because the free protein has sections missing in the pdb.
Because the free protein structure has residues missing presumably due to
disorder in the crystal, we chose to use the structure of the bound
protein. Could the use of a structure of our protein with more order than
expected cause problems with the selection of the rotational diffusion
tensor?
Very much so! If there is a conformational change, you should
probably split the protein in 2 and analyse each domain or rigid body
separately.
The diffusion tensor the program has selected doesn’t seem to match the
overall shape of the protein. If it can cause a problem, how do I add the
residues missing from the crystal structure for the refinement?
Furthermore, can the program handle some residues missing in crystal
structures?
The problem with a non-spherical model-free analysis is that the
average XH bond vector orientation in solution is a core requirement
of the theory. If this is incorrect, this can lead to either of the
two well known artefacts:
- Artificial nanosecond motions (the Schurr 1994 paper).
- Artificial Rex value (the Tjandra 1995 paper).
See my 2007 Mol. Biosyst. paper at http://dx.doi.org/10.1039/b702202f
for a review about this problem and how this relates to the under or
overestimation of the vector projection within the diffusion tensor.
One way to handle this situation is to use what I have termed a
'hybrid' model. For residues which have no structural information,
you can use the local tm model and avoid a global correlation time,
and linearly combine these models with the global diffusion tensor
models applied to all known residues. You could even shift residues
with high beta factors from the global Brownian diffusion model into
the 'local tm' model category. relax will allow you to do this, but
it will require a bit of manual work. So if you know you have two
domains with large structural rearrangements, you can perform three
analyses:
- Local tm for missing residues.
- The full protocol for domain 1.
- The full protocol for domain 2.
Then combine the final results as one model. There should be zero
overlap of residues in these 3 divisions. You could then compare
this, using AIC values to take parsimony and non-nested model
differences into account, to a local tm model for all residues. Note
that the local tm models are generally quite noisy if data from < 3
fields are used. I hope this helps.
Regards,
Edward
Re: Problems Loading PDB