CROSS-VALIDATED MAXIMUM-LIKELIHOOD ENHANCES CRYSTALLOGRAPHIC SIMULATED ANNEALING REFINEMENT

Recently, the target function for crystallographic refinement has been improved through a maximum likelihood analysis, which makes proper al lowance for the effects of data quality, model errors, and incompleten ess, The maximum likelihood target reduces the significance of false l ocal minima during the refinement process, but it does not completely eliminate them, necessitating the use of stochastic optimization metho ds such as simulated annealing for poor initial models, It is shown th at the combination of maximum likelihood with cross-validation, which reduces overfitting, and simulated annealing by torsion angle molecula r dynamics, which simplifies the conformational search problem, result s in a major improvement of the radius of convergence of refinement an d the accuracy of the refined structure, Torsion angle molecular dynam ics and the maximum likelihood target function interact synergisticall y, the combination of both methods being significantly more powerful t han each method individually, This is demonstrated in realistic test c ases at two typical minimum Bragg spacings (d(min) = 2.0 and 2.8 Angst rom, respectively), illustrating the broad applicability of the combin ed method, In an application to the refinement of a new crystal struct ure, the combined method automatically corrected a mistraced loop in a poor initial model, moving the backbone by 4 Angstrom.