Electronic-structure-based molecular-dynamics method for large biological systems: Application to the 10 basepair poly(dG)center dot poly(dC) DNA double helix

Combining several recently developed theoretical techniques, we have developed an electronic-structure-based method for performing molecular-dynamical simulations of large biological systems. The essence of the method can be summarized in three points: (i) There are two energy scales in the Hamiltonian and each is treated differently-the strong intramolecular interactions are treated within approximate density-functional theory, whereas the weak intermolecular interactions (e.g., hydrogen bonds) are described within a simple theory that accounts for Coulomb, exchange, and hopping interactions between the weakly interacting fragments. (ii) A localized basis of atomic states is used, yielding sparse Hamiltonian and overlap matrices. (iii) The total energies and forces from the sparse Hamiltonian and overlap matrices are solved using a linear scaling technique to avoid the N-3 scaling problem of standard electronic structure methods. As an initial benchmark and test case of the method, we performed calculations of a deoxyribonucleic acid (DNA) double-helix poly(dG). poly(dC) segment containing ten basepairs, with a total of 644 atoms. By a dynamical simulation, we obtained the minimum-energy geometry and the electronic structure of this DNA dehydrated segment, as well as the full dynamical matrix corresponding to the relaxed structure. The vibrational data and energy band gap obtained compare qualitatively well with previous experimental data and other theoretical results.