Map‐likelihood phasing

The recently developed technique of maximum‐likelihood density modification [Terwilliger (2000), Acta Cryst. D 56 , 965–972] allows a calculation of phase probabilities based on the likelihood of the electron‐density map to be carried out separately from the calculation of any prior phase probabilities. Here, it is shown that phase‐probability distributions calculated from the map‐likelihood function alone can be highly accurate and that they show minimal bias towards the phases used to initiate the calculation. Map‐likelihood phase probabilities depend upon expected characteristics of the electron‐density map, such as a defined solvent region and expected electron‐density distributions within the solvent region and the region occupied by a macromolecule. In the simplest case, map‐likelihood phase‐probability distributions are largely based on the flatness of the solvent region. Though map‐likelihood phases can be calculated without prior phase information, they are greatly enhanced by high‐quality starting phases. This leads to the technique of prime‐and‐switch phasing for removing model bias. In prime‐and‐switch phasing, biased phases such as those from a model are used to prime or initiate map‐likelihood phasing, then final phases are obtained from map‐likelihood phasing alone. Map‐likelihood phasing can be applied in cases with solvent content as low as 30%. Potential applications of map‐likelihood phasing include unbiased phase calculation from molecular‐replacement models, iterative model building, unbiased electron‐density maps for cases where 2 F o − F c or σ A ‐weighted maps would currently be used, structure validation and ab initio phase determination from solvent masks, non‐crystallographic symmetry or other knowledge about expected electron density.