Simon M. Brocklehurst (smb at bioch.ox.ac.uk) wrote:
: Shoba.Ranganathan wrote:
: >
: > Hi,
: >
: > I'm looking for references that state the minimum sequence identity
: > required for homology modelling, and the percentage in the reference
: > (in case I can't find the article).
: >
: You might want to consider using the term "comparative modelling", as opposed
: to "homology modelling". The reason for this is that the minimum sequence
: identity that is "required" for useful 3-D prediction and modelling can be as
: low as zero percent (or some other statistically insignificant
: similarity). Thus it is, in fact, possible to model the structure of a protein
: based on the structure of an unrelated (or very distantly related) protein.
: | ,_ o Simon M. Brocklehurst,
I think Simon is confusing the issue here. Sure, unrelated proteins can
adopt the same fold, but I think the question related to standard
homology modelling techniques. If your sequence identity is above the
30% level, you can be pretty certain that the proteins adopt the same fold
and would therefore chose a standard homology modelling procedure whether
it be manual, or automated (e.g. Composer or Modeller). The higher the
homology, the more accurate your model is likely to be. Below the 30%
level, you have to use fold recognition tecniques such as threading
(e.g. THREADER) or secondary structure prediction / pattern matching
(e.g. MAP).
Coming onto the question itself, around 30% is the generally quoted
figure, but I can't think of any references off-hand. You might take
a look at papers by Tom Blundell, Janet Thornton, Christine Orengo, and
Andre Sali for guidance.
Andrew
--
****************************************************************************
Dr. Andrew C.R. Martin, University College London & SciTech Software
EMAIL: martin at biochem.ucl.ac.uk Tel:(Work) +44(0)171 419 3890
URL: http://www.biochem.ucl.ac.uk/~martin (Home) +44(0)1372 275775
****************************************************************************
