Automated analysis of large sets of heteronuclear correlation spectra in NMR-based drug discovery.

摘要

Drug discovery procedures based on NMR typically require the analysis of thousands of NMR spectra. For example, in "SAR by NMR", two-dimensional NMR spectra are recorded for a target protein mixed with ligand candidates from a comprehensive library of small molecules and are compared to the corresponding spectrum for the protein alone. We present an automated procedure for the comparative analysis of large sets of heteronuclear single quantum coherence spectra, which is based on three-way decomposition and implemented as the software package MUNIN. In a single step, spectra with differences in the peak positions (indicating ligand binding) and the affected peaks are identified. By omission of peak picking, ad hoc scoring of the quality of doubtful peaks is avoided. The procedure has been tested on the bacterial ribonuclease barnase, with a protein concentration of only 50 microM, using several small molecules including the substrate analogue 3'-GMP. Sets of 51 spectra were processed simultaneously, and it is concluded that spectra with binding ligands can be unambiguously identified from much larger sets of spectra.