In article <9501041949.AA28566 at silibone.cchem.berkeley.edu>,
yang at SILIBONE.CCHEM.BERKELEY.EDU wrote:
>> We are looking for programs which search backbone structures for metal-binding
> sites--specifically, we are interested in "designing" such sites in
> non-metallo enzymes in order to improve catalysis. Thanks in advance!
You're probably aware of WHATIF by G.Vriend - if not, check Vriend,
J.Mol.Graph, 8:52-56, and Proteins Struct.Funct. Genet. 11:52-58. To obtain
a licence, contact <vriend at embl-heidelberg.de>. This should help with
identifying binding sites.
The question of how to put ligand binding sites into your protein is a
different matter. You need to identify those residues of your target
structure, which are closest in 3D structure to the binding site. But you
know in advance neither where these residues are going to be, nor even in
what order they should be in the target sequence. For example, a three His
zinc binding site may be -1-..-2-..-3- in the binding site from the
database, while the closest structural correspondence in your target
sequence might be -2-..-1-..-3- and with the residues on totally different
elements of secondary structure.
What this amounts to is a combinatorial search of _all_ possible
replacements and then checking which replacement will give the best fit.
Then you can find those which are plausibly close to your active site among
well fitting candidates. The problem is of course that the number of tests
you have to do explodes (it is n!-(n-i)! with n the size of your protein
sequence and i the size of your binding site, which for a site of 6
residues tested against a protein of 200 residues is 6*10^13) A while ago I
wrote a program to do this kind of test efficiently. Contact me if you need
Boris <steipe at lmb.uni-muenchen.de>