Скачать или смотреть CCP4 Study Weekend 2019 - Information content in MR - Randy Read

Likelihood-based methods have come to play a prominent role in macromolecular crystallography. For molecular replacement (MR), likelihood allows structures to be solved with models from more distant relatives or with fragments that comprise smaller fractions of the asymmetric unit1. Understanding the properties of the likelihood function, and of the likelihood scores that are expected to be achieved in a particular search, allow us to predict the outcome of MR searches2 and to make rational decisions about strategy3.
The success of likelihood-based methods play suggests that it is worth examining the closely-related concept of information. The process of carrying out a diffraction experiment yields information about the true values of the diffraction intensities; can a measure of the information added by the diffraction experiment provide a useful criterion to test whether we have collected actual data and not just hkl values? How is this influenced by the presence of anisotropy and translational non-crystallographic symmetry? Can information measures define the overall resolution limit for useful data? Does the information content of the data determine the number of parameters that we can estimate?
Applications of these information concepts to molecular replacement calculations will be addressed.