A SIMPLE METHOD FOR THE REFINEMENT OF MODELS DERIVED FROM NMR DATA DEMONSTRATED ON A ZINC-FINGER DOMAIN FROM YEAST ADR1

A computationally facile and easily implemented iterative method for refining structures derived from NMR data is presented. This method combines a distance geometry algorithm (DIANA), a complete relaxation-matrix algorithm (CORMA), and a new algorithm that uses the sixth root of the ratio between observed and calculated NOE intensities as a refinement gradient (REPENT). As demonstrated here on a zinc-finger peptide, the REPENT protocol converges rapidly to a low residual error. Comparison of the REPENT-refined models and models generated using a commonly employed method of assigning NOEs as strong, medium, or weak reveals that the refined models are superior in their level of definition, their ability to reproduce the experimentally observed data, and their ability to predict salient structural features. © 1993 Academic Press, Inc.