On Nov 2 2007, Bryan W. Lepore wrote:
neglecting e-density, packing, etc. - what matters most to people when running phaser - LLG or Z-score? in what function (RF, TF)?
Depends on the purpose. From the LLG, I want to see that it is positive (negative means that I'm being too optimistic about the quality of the model, i.e. the RMS error is higher or the completeness lower than assumed), and I would like to see it increase as the solution becomes more complete (i.e. RF score increases with TF, which increases for RF on second molecule, and so on). But to assess the significance of a solution, I'd place more weight on the LLG. For various reasons, the RF can be unclear, particularly when there is high symmetry or high NCS, so a correct orientation can have a low Z-score. But usually the Z-score for a correct translation is a better indication. In some cases with multiple molecules in the a.u., the Z-score gets higher for molecules later in the search.
further, if you 'refine' the RMSD to find the best RMSD, what is the more robust indicator - LLG or Z-score?
LLG is the one that tells you how well your model fits the data, and the assumed RMSD is one of the parameters of the model. At some point, we will refine the assumed RMSD in Phaser (against the LLG), but the signal can be very weak for incomplete models, so we'll have to be careful about how we implement that.
lastly - is it better to have a high LLG/Z-score with poor discrimination from other peaks or the other way around?
Zscore and discrimination tend to be highly correlated, and we don't really have a good feel for what different absolute values of LLG mean. Anyway, it's the discrimination from other peaks that convinces you that the top solution is the correct one, and not one of the other solutions. If there's poor discrimination, none of the solutions is correct or you need some additional information to sort them out. This can be things like how well they behave in subsequent refinement, or whether the model can be used to find sensible positions for anomalous scatterers.
Randy Read
