STRUCTURED MATRIX PERTURBATIONS FOR MOLECULAR CONFORMATIONS

摘要

Three-dimensional molecular structure is fundamental in drug design and discovery, docking, and chemical function identification. The input to our algorithm consists of a set of approximate interatomic distances or distances constrained in intervals of varying precision; some are specified by the covalent structure and others by NMR experiments and application of the triangle inequality. The output is a valid molecular conformation in a specified neighborhood of the input. We aspire that our approach helps in detecting outliers of the NMR experiments, and that it manages to handle partial inputs. Numerical linear algebra methods are employed for reasons of speed and accuracy. The main tools include, besides iterative local optimization, distance geometry and matrix perturbations for minimizing singular values of real symmetric matrices. Our algorithm is able to bound the number of degrees of freedom on the conformation manifold. A public domain MATLAB implementation is described; it can determine a conformation of a molecule with 20 backbone atoms in 3.79sec on a 500MHz PENTIUM-Ⅲ.