A new method for structure refinement, based on optimization of the variable target function (Sugár and Xu, Prog. Biophys. Mol. Biol.58, 61, 1992), is examined in order to explore the conditions of high-resolution polypeptide structure determination. The method of variable target function was used to analyze simulated NOESY spectra of different spectral resolutions, of different signal/noise ratios, and of different number of mixing times. The goodness of the results was assessed by comparing the known polypeptide conformation with the optimized ones, as well as by comparing the simulated experimental spectra with the spectra back-calculated from the optimized structures. According to the test calculations, in order to get high-resolution polypeptide structures, the number of constraints should be larger than the number of dihedral angles, and the constraints should be distributed over the dihedral angles more or less homogeneously. The success rate of the variable-target-function method in finding high-resolution structures was high until a critical signal/noise ratio, while below this threshold the success rate was low. The critical signal/noise ratio could be decreased, however, by increasing the number of mixing times. The test calculations show that the variable-target-function method is capable of high-resolution polypeptide structure determination with a success rate of 85-90% under common experimental conditions, such as 10 constraints/residue, 40% noise/signal ratio, and six mixing times. © 1995 Academic Press, Inc.