The proposed research focuses on a multidisciplinary approach toward computation of macromolecular structure. The aim is to (1) develop new algorithms for multivariate minimization and molecular dynamics on supercomputers, and (2) apply them to outstanding problems in nucleic acid and protein dynamics. A key vehicle is the Langevin/Implicit-Euler scheme for molecular dynamics, permitting large time steps and damping high-frequency modes. Efficient conformational searches and global optimization strategies will be applied to important folding problems in DNA and proteins.