NAME: Phil Evans

AFFILIATION: MRC Laboratory of Molecular Biology, UK

CONTACT:

 

TITLE: "Introduction to Molecular Replacement"

ABSTRACT: In principle, molecular replacement is very simple. We have diffraction data from a crystal of an unknown structure, and a known related structure which we believe to be a good model. Then in the computer we place the model in all possible orientations and translations, at each point score how well the predicted diffraction pattern matches the observations, and pick the best score. In practice, each stage has problems.

 

• how to choose the best model: this will be discussed in another session, but it is the most important stage. When molecular replacement fails, it is always because the model is not sufficiently like the true structure.

• what is the best score to use? The traditional scoring methods are based on Patterson superposition, using a product function. Maximum likelihood methods allow for the errors in both model and experimental data, and give better discrimination between true and false solutions.

• What is the best strategy for searching all possible orientations and translations? An exhaustive search of 3 rotations and 3 translations can take a very long time, even with modern computers: it is even slower if there are more than one molecule in the asymmetric unit (6N parameters). To speed up the calculation, the search is usually split into a rotation search followed by a translation search. The way in which rotations are described, and the search volume for translations, will be discussed.

 

A number of computational tricks are used to make the calculations faster, principally by factorising the structure factor expression: these will be described in outline.