I agree with Carlos's suggestions (search for your two domains
separately, and as poly-Ala models).
There's a few other things that may help MR, which you may or may not
have tried:
1. Reset the B-factor of your search models to the wilson B-factor of
your dataset.
2. Search using only lower-resolution data (aka if you've got 2.5 - 50,
search using 3.5-30 or 4.0-30).
3. If you've got homologous structures for your domains (it sounds like
you do), align them and remove any regions that vary significantly from
your search model. Model inaccuracy will cause worse problems than
model incompleteness, at least in my experience.
4. Search in all alternative spacegroups - it adds a little to the
runtime, but IMHO is worth it.
Pete
RHYS GRINTER wrote:
Hi All,
I'm currently working on solving the structure of a protein by molecular
replacement. The protein is around 30kDa and likely has a two beta-prism
domains, linked by a long curved two stranded sheet based on the structure of
an analogue. There are also a number of other structures which represent a
single homologous beta-prism domain.
I've tried to find MR solution using the analogue and various truncation/AA
substitution models based on it with no success. I've also tried single domain
ensembles of the other homologous structures, also with no success. I think the
problem is the overall sequence homology is quite low between my protein and
the available structures (35% for the analogue and around 20% for the other
models.
I was curious as to how someone with more experience would tackle this problem.
Just for background, the datasets I have are 2 to 2.7 angstroms with pretty
nice stats. The space group is most likely C2221 with two molecules per ASU
(giving around 58% solvent).
Thanks,
Rhys Grinter
PhD Candidate
University of Glasgow
