AUTOMATED PROCESSING OF 2-DIMENSIONAL CORRELATION SPECTRA

An automated scheme is described which locates the centers of cross pe aks in two-dimensional correlation spectra, even under conditions of s evere overlap. Double-quantum-filtered correlation (DQ-COSY) spectra h ave been investigated, but the method is also applicable to TOCSY and NOESY spectra. The search criterion is the intrinsic symmetry (or anti symmetry) of cross-peak. multiplets. An initial global search provides the preliminary information to build up a two-dimensional ''chemical shift grid.'' All genuine cross peaks must be centered at intersection s of this grid, a fact that reduces the extent of the subsequent searc h program enormously. The program recognizes cross peaks by examining the symmetry of signals in a test zone centered at a grid intersection . This ''symmetry filter'' employs a ''lowest value algorithm'' to dis criminate against overlapping responses from adjacent multiplets. A pr ogressive multiplet subtraction scheme provides further suppression of overlap effects. The processed two-dimensional correlation spectrum r epresents cross peaks as points at the chemical shift coordinates, wit h some indication of their relative intensities. Alternatively, the in formation is presented in the form of a correlation table. The authent icity of a given cross peak is judged by a set of ''confidence criteri a'' expressed as numerical parameters. Experimental results are presen ted for the 400-MHz double-quantum-filtered COSY spectrum of 4-androst en-3,17-dione, a case where there is severe overlap. (C) 1998 Academic Press.