Statistical strategy for stereospecific hydrogen NMR assignments: validation procedures for the floating prochirality method.

We examine the statistical and other considerations which determine the validity and reproducibility of stereospecific hydrogen NMR assignments obtained by the floating prochirality method. In this method, the assignment of a prochiral configuration of hydrogens at selected centers is allowed to 'float' during the structure refinement, and the distribution of prochiral orientations in highly refined structures is subjected to statistical analysis. The underlying statistical basis for this approach is examined and potential limitations of current approaches are identified. As an example, approximately 1300 distance constraints obtained from NOESY spectra of oxidized horse cytochrome c have been used to examine several computational strategies. Repeated calculations were done by several different methods on both the whole molecule (104 residues plus heme) and on a 23-residue fragment containing two helices, a turn, and flanking residues. The results show that, even with NOE constraints alone, one third of the centers may be reproducibly assigned, provided appropriate precautions are taken. These precautions include adjustments for multiple statistical comparisons and characterization of statistical interactions between prochiral centers. The analysis demonstrates that inadequately constrained systems, such as fragments from a larger molecule, may produce misleading results, raising concerns about methods which rely solely on intraresidue and sequential interresidue constraints. A mathematical model describing interactions among prochiral centers is described and validated, and protocols for assignment and statistical validation are presented.