GARANT-a general algorithm for resonance assignment of multidimensional nuclear magnetic resonance spectra

A new program for automatic resonance assignment of nuclear magnetic Ž . Ž resonance NMR spectra of proteins, GARANT General Algorithm for . Resonance AssignmeNT , is introduced. Three principal elements used in this Ž . approach are: a representation of resonance assignments as an optimal match of two graphs describing, respectively, peaks expected from combined knowledge of the primary structure and the magnetization transfer pathways in Ž . the spectra used, and experimentally observed peaks; b a scoring scheme able Ž . to distinguish between correct and incorrect resonance assignments; and c combination of an evolutionary algorithm with a local optimization routine. The score that evaluates the match of expected peaks to observed peaks relies on the agreement of the information available about these peaks, most prominently, but not exclusively, the chemical shifts. Tests show that the combination of an evolutionary algorithm and a local optimization routine yields results that are clearly superior to those obtained when using either of the two techniques separately in the search for the correct assignments. GARANT is laid out for assignment problems involving peaks observed in twoand three-dimensional homonuclear and heteronuclear NMR spectra of proteins. Q 1997 by John Wiley & Sons, Inc. *Author to whom all correspondence should be addressed. ( ) Journal of Computational Chemistry, Vol. 18, No. 1, 139]149 1997 Q 1997 by John Wiley & Sons CCC 0192-8651 / 97 / 010139-11