Torsion-angle molecular dynamics as a new efficient tool for NMR structure calculation.

Molecular dynamics in torsion-angle space was applied to nuclear magnetic resonance structure calculation using nuclear Overhauser effect-derived distances and J-coupling-constant-derived dihedral angle restraints. Compared to two other commonly used algorithms, molecular dynamics in Cartesian space and metric-matrix geometry combined with Cartesian molecular dynamics, the method shows increased computational efficiency and success rate for large proteins, and it shows a dramatically increased radius of convergence for DNA. The torsion-angle molecular dynamics algorithm starts from an extended strand conformation and proceeds in four stages: high-temperature torsion-angle molecular dynamics, slow-cooling torsion-angle molecular dynamics, Cartesian molecular dynamics, and minimization. Tests were carried out using experimental NMR data for protein G, interleukin-8, villin 14T, and a 12 base-pair duplex of DNA, and simulated NMR data for bovine pancreatic trypsin inhibitor. For villin 14T , a monomer consisting of 126 residues, structure determination by torsion-angle molecular dynamics has a success rate of 85%, a more than twofold improvement over other methods. In the case of the 12 base-pair DNA duplex, torsion-angle molecular dynamics had a success rate of 52% while Cartesian molecular dynamics and metric-matrix distance geometry always failed.