Improvement in accuracy of protein local structure determination from NMR data

Abstract

A method for determining the most probable conformations of amino acid residues from semiquantitatively estimated nuclear Overhauser effects (NOEs) and coupling constants was developed and coded in the FiSiNOE-2 program. This program is a new version of the FiSiNOE program, utilizing NMR data with complementary knowledge-based information on protein structures. In FiSiNOE-2 this information is conformational clusters of the dihedral angles (φ, ψ, gc 1) derived from the Protein Data Bank. The FiSiNOE-2 method determines mathematical expectations and standard deviations for the angles φ, ψ, and χ 1, and provides direct determination of the local structure of proteins from NMR data before building and refining their spatial structure. The results of the FiSiNOE-2 program in combination with the results of the habas program may be used to provide stereospecific assignments of a pair of β-methylene protons and to determine precisely allowed ranges of the φ, ψ, and χ 1 dihedral angles consistent with a given set of NMR data. To do this, a new procedure, combine, was developed. Computational experiments with the NMR data simulated from X-ray coordinates of the BPTI showed that use of the combine procedure, in comparison with results obtained when habas was used alone, increases by more than 30% the number of correct assignments for βCH 2 groups and reduces the total lengths of the combined angular intervals for φ, ψ, and χ 1 angles to 1.9, 2.4, and 1.8 times, respectively. In contrast to the redundant dihedral angle constraints (REDAC) strategy, that derives REDAC from preliminary calculations of the complete structure, the combine procedure reduces the length of the angular intervals before using the variable target function algorithm to determine spatial structures of proteins. This feature of the combine strategy may be especially beneficial in the cases when there is lack of long-range NOEs.