Page 116 - Macromolecular Crystallography

MOLECULAR REPLACEMENT TECHNIQUES 105

efforts should be put into the generation of the best deposited in the PDB, which usually contain 20
possible model. a priori ‘equivalent’ models, specialized protocols
have been devised and tested (Chen, 2001), some of
which are available on Gerard Kleywegt’s web site
7.8.1 The best starting model
(http://xray.bmc.uu.se/usf/factory_6.html). One
Choosing the best model, even when only one pos- idea is to use an ‘average’ structure; another idea is
sible homolog has been detected through alignment to weight each atom by a pseudo B-factor, which is
and/or threading methods, is not an easy task. This calculated by an empirical formula that is a function
is usually dealt with by running MR, in turn, with of the RMSD of the position of this atom in the 20
different versions of the same model. Up to now, the different models of the PDB (see Protocol 7.5). This
general accepted rule of thumb was to remove those usually works well (Wilmanns and Nilges, 1999).
parts of the model that are suspected to be different
in the target protein. In other words, it is believed 7.8.2 Using homology-modelling derived
to be better to have an incomplete model with no models
error than a complete one with errors. For instance,
one could truncate all side chain atoms downstream Up to now, models of the protein derived by
of the CB (except for glycine) effectively changing homology-modelling techniques were not heavily
the sequence into a poly-Ala; or one could choose to used, because people were reluctant to use models
keep the side-chain atoms coordinates of only those which contain some errors. In these models, all side-
residues that are strictly conserved between the tem- chains have been reconstructed, as well as insertions
plate and the target and mutate all the others into and deletions. So, in a sense, the model is more com-
alanine; or into a serine (changing the CG atom into plete, but it is not certain that this will facilitate the
an OG atom) if there is a conservative substitution. search for the solution of the molecular replacement
There are all sorts of possibilities, including solution. One fearful feature of homology modelling
the one to keep all atoms inside the core of the is that the refinement of the model using standard
molecule and truncating atoms with an accessi- force-field tends to worsen the model, rather than
bility to the solvent larger than a given criterion improve it, at least in test cases. Also, as there are
(Delarue et al., 1990). Truncating loops with high several homology modelling programs available on
B-factors or non-conserved loops is also possible, the web, all of which use different methods (distance
although it is only recently that automatic pro- geometry: Modeller (Sali and Blundell, 1993), mean-
tocols have been devised to do the latter, using field optimization techniques (Koehl and Delarue,
information contained in the multialignement of the 1994, 1995)) and sometimes depend on the genera-
sequences (Claude et al., 2004). The richest diver- tion of random numbers, the question arises as to
sity of approaches is the one available in MrBump which should be used.
(Keegan and Wynn, 2007). When there are several It is well known that the result of homol-
possible models (templates), the situation becomes ogy modelling is highly dependent on the qual-
more difficult to handle; if there are only two or ity of the alignment between the template and
three, they can be tried individually, say as poly- the target. Obviously, the success of MR will be
Ala models. If there are more, as in structures highly dependent on the accuracy of the align-
solved by Nuclear Magnetic Resonance (NMR) and ment between the template and the target sequences

Protocol 7.5 How to handle NMR models (Chen, 2001; Wilmanns and Nilges, 1999)
Go to Gerard Kleywegt’s website (Uppsala Software Get the script in:
Company): ftp://xray.bmc.uu.se/pub/gerard/omac/multi_probe
http://xray.bmc.uu.se/usf/factory_6.html and follow the advice given by Y. W. Chen (2001).