Predictions of protein backbone bond distances and angles from first principles

A procedure is described, based on a spline‐function representation of ab initio peptide conformational geometry maps, that allows one to predict backbone bond distances and angles of proteins as functions of the peptide ϕ(N‐Cα)/Ψ(Cα‐C′) torsions with an accuracy comparable to that of high‐resolution protein crystallography. For example, for the more than 40 residues of crambin, the rms deviation between predicted and crystallographic values of N‐Cα‐C′ is 1.9° for the 1.5 Å resolution structure and 1.9° for the 0.83 Å resolution structure, compared with angle variations of < 10°. Accurate information on protein backbone geometries is important for establishing dictionaries of flexible geometry functions for use in empirical peptide and protein modeling.