Professor Ponder develops and applies computational tools for problems in structural biology and in protein engineering, function, and folding. His group's research focuses on prediction and modeling of structural chemistry and the relation of structure to molecular properties.
The Ponder group develops and applies computational tools for problems in structural biology and in protein engineering, function and folding. The Ponder Lab produces and distributes software packages ranging from macromolecular mechanics and dynamics simulation (TINKER) to molecular visualization (Force Field Explorer) to empirical packing analysis of protein structure (PROPAK) to sequence analysis and tertiary structure prediction (SLEUTH). Our research focuses on prediction and modeling of structural chemistry and the relation of structure to molecular properties.
Our major research area in recent years has concerned implementation of efficient methods for including multipole electrostatics and polarization in simulations as a framework for our next-generation AMOEBA force field. This new energy model enables reliable calculation of structures and has significant advantages over traditional fixed partial atomic charge models such as Amber and CHARMM. It also yields energetics for ligand docking and drug design to within "chemical accuracy"-- absolute errors of 0.5 kcal/mol or less. Current AMOEBA applications include free energy calculations of binding interactions, elucidation of the role of ions in biology, prediction of small molecule crystal structures and refinement of highly accurate protein homology models. Planned future improvements to the AMOEBA model are ligand field splitting terms for transition metals, incorporation of charge transfer and penetration effects, improved treatment of short-range dispersion, and coupling of valence geometry to electrostatics.
In addition, we are exploring various powerful approaches to conformational search for flexible biopolymers. One method transforms the potential energy surface for a molecule by a diffusion equation-based smoothing procedure. This "potential smoothing" paradigm is applicable to a variety of problems including transmembrane helix packing, global optimization, and energy-based clustering of conformations. Another search method is based on a novel distance geometry algorithm and heuristic rules as a basis for protein structure prediction. Statistical distance distributions and predicted secondary structure constraints generate libraries of candidate folds to be scored with an informatics-based contact function or physics-based effective mean force potential. Ultimately, our interest in conformational search lies in the "end game" of protein folding--in making a connection between atomic-level protein structures and low-resolution models available from fold recognition algorithms.