Premium
MORMIN: A quasi‐Newtonian energy minimizer fitting the nuclear overhauser data
Author(s) -
Pothier JoëL,
GabarroArpa Jacques,
Bret Marc Le
Publication year - 1993
Publication title -
journal of computational chemistry
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.907
H-Index - 188
eISSN - 1096-987X
pISSN - 0192-8651
DOI - 10.1002/jcc.540140210
Subject(s) - eigenvalues and eigenvectors , nuclear overhauser effect , cartesian coordinate system , exponential function , function (biology) , scalar (mathematics) , quadratic equation , mathematical analysis , mathematics , physics , nuclear magnetic resonance , quantum mechanics , geometry , nuclear magnetic resonance spectroscopy , evolutionary biology , biology
In this article, we describe the program MORMIN, which can simultaneously minimize the mechanical energy of a given macromolecular structure, together with a weighted quadratic penalty function of the difference between the observed and computed nuclear Overhauser effect (nOe) peaks. The gradient of the nOe penalty function relatively to the proton coordinates is computed from an exact closed formula of a matrix exponential derivative. To cut CPU time, the molecular system is partitioned into nonoverlapping subsets containing the protons involved in the observed peaks. The algorithm is no longer exact, but if a 1% relative error is accepted it can be run, on a scalar computer, in about the same CPU time as needed for the calculation of the mechanical energy. We have successfully run the program in more than 1000 situations, including cases where the hybrid method failed because of the occurrence of negative eigenvalues. In some cases, the optimization of the Cartesian coordinates could be successfully extended to individual atomic diffusion times. © 1993 John Wiley & Sons, Inc.