Re: CCL:predict 3D structure from sequence (fwd)

Dear Csaba and all,
 "Dr. Csaba Hetenyi" wrote:
 > > The first one called Camparative Protein Modeling, which requires a 3d
 > > structure of a protein, which has >30% sequence Similarity
 comparing to
 > > target. It's pretty reliable.  The second approach has lower success
 > > rate..
 >
 > I'd rather not say "it is pretty reliable". In my view, sequence
 > similarity (30 % or above) is not a sufficient only a necessary
 > requirement of possible homology modeling (HM).
 > A "nice" example of failure of HM was recently mentioned in
 Williams et
 > al. Science 305 (2004) 683-686. E.g. CYPs are "excellent"
 examples of
 > proteins not appropriate for HM.
 According to your own necessary/sufficient conditions this wouldn't be a
 good
 example, since the authors state that templates available for
 comparative modeling
 of CYP3A4 display just sequence identities < 30% with the target. On p.
 684 you
 can read:
 "The most widely used templates for modeling CYP3A4 are the P450
 structures from
 Bacillus megaterium (P450 BM3) and Saccharopolyspora erythraea (P450
 EryF), both
 of which share less than 25% sequence identity with CYP3A4."
 Without the intension of self-advertising, I can refer you to our
 (Hillisch, A.,
 Pineda, L.F. and Hilgenfeld, R.) recent paper published on DDT
 9(2004)659-669 on
 the utility of homology models in the drug discovery process, wich
 contains a, in
 our opinion, balanced discussion on the performance of the comparative
 modeling
 method in different applications.
 Best regards
 Felipe