Computer-Assisted Structure Determination of Biomacromolecules by NMR

Abstract

In the past years the use of multidimensional heteronuclear techniques has pushed the limit in size of the molecular systems studied by NMR well beyond 100 amino acids (Wagner 1993). The resulting growth in complexity of the NMR spectra places an increasingly important role on computer programs that support the interpretation of the spectra and the management of the assignment data. When the earliest NMR structure determinations were accomplished about ten years ago (Wüthrich 1986), this was not evident for the sequence-specific assignment of the spectral cross peaks (Billeter et al. 1982). In contrast, the distance geometry calculations were always performed in an almost exclusively automated manner (Havel and Wüthrich 1984). Yet, still today two different uses of computers are observed in NMR structure determinations. One consists of the interactive process of obtaining cross peak assignments and extracting structural information from the NMR spectra with the help of a computer graphics package running on a workstation. The other typically consists of the submission of lists with structural constraints to a “black box”, e.g. a program running on a remote supercomputer, in order to obtain a three-dimensional structure. Both approaches are subject to possible errors. The wealth of spectral information is difficult to interpret in a consistent way, and distance geometry approaches are known to be biased in various ways.