B. Strandberg

The method of multiple isomorphous replacement has been used to solve the structure of the protein myoglobin to a resolution of 2 A. EDSAC II has been programmed to calculate the phase angles, and the statistical treatment of errors proposed by Blow & Crick has been used throughout, leading not to the most probable electron-density map but to the one with the least mean square error over the entire unit cell. Problems inherent in adapting the process for a digital computer are discussed, and examples are given of probability curves for typical phase determinations. A method of refining heavy-atom parameters in the course of phase determination is presented. Comparisons of the mostprobable and least-error electron-density maps are presented in the region of the haem group, and it is shown that the latter map gives slightly better results. In an appendix a means of trial-anderror least-squares refinement of heavy-atom parameters using data from a centrosymmetric projection is discussed.