Interdisciplinary Initiatives Program Round 2 – 2002

Jean-Claude Latombe, Computer Science
Patrice Koehl, no longer at Stanford
Michael Levitt, Structural Biology
Leonidas Guibas, Computer Science
Eric Darve, Mechanical Engineering

Computational structural biology requires exploring conformational spaces of proteins, e.g., to study thermodynamic properties, find native folds, and generate decoy sets. These spaces have many dimensions—typically, in the hundreds, or more. The goal is often to find low-energy conformations. But energy functions involve many terms, most of which (electrostatic, van der Waals, and heuristic potentials) contributed by atom pairs. We develop new data structures and algorithms to efficiently explore protein conformational spaces and maintain energy values. The underlying idea is to exploit the unique geometric, kinematic and physical properties of proteins—e.g., they form long kinematic backbones with short side-chains, most molecular forces are short-range—to speed up computations.