Parameter refinement in the multiple isomorphous-replacement method

The usual procedure for parameter refinement in the multiple isomorphous-replacement method leads to a heavy bias in each round of refinement towards the previously assumed values. This leads to difficulties of convergence, which can be severe when one derivative or site is overwhelmingly powerful in phase determination. The bias arises from using the parameters of a derivative in estimating phase angles, which are then used to refine these same parameters. By omitting the derivative which is to be refined from the phase determination, this bias is avoided, and even though the phases may then be very inaccurate, convergence is rapid. This method was applied to the α-chymotrypsin structure, when the conventional procedure failed to converge correctly, and led to a greatly improved electron-density map. Correct estimation of occupancy and lack-of-closure errors remains an unsolved problem.