LANS Informal Seminar: Jaydeep Bardhan

Seminar Title: The Devils in the Details: Numerical Algorithms for Molecular Analysis and Design
Speaker: Jaydeep Bardhan, MCS

Date/Time: 2008-02-21 15:00
Location: Building 221, Room A-261

Description:
For decades, theoretical and computational scientists at the intersection of biology, chemistry, and physics have demonstrated the power of using computation to study and modify molecular interactions. Many of the theoretical models for these interactions are highly approximate–usually by necessity–and it is important to characterize their strengths and weaknesses in a variety of contexts. The availability of efficient numerical algorithms can therefore significantly improve our ability to assess and compare models, a task of increasing importance as we look to engineering molecular systems in biology and in general. This talk will describe some of our work developing numerical algorithms related to solvation phenomena in biology–how water and small concentrations of ions affect molecular structure and function. First, a collaborative effort at the Advanced Photon Source focuses on characterizing wide-angle X-ray scattering experiments of proteins in solution. These studies offer exciting possibilities for improving our understanding of the relation between protein dynamics and function. A second area of research focuses on continuum models commonly used to study electrostatic interactions between biomolecules and the surrounding solvent. We have been developing numerical methods to improve the accuracy with which these models can be solved using boundary-element methods. One interesting result suggests a new interpretation for Generalized-Born (GB) treatments of electrostatic interactions. Finally, in the context of molecular design, we will discuss fast algorithms that we have developed for an electrostatic-optimization framework for analyzing and improving molecular interactions. These algorithms, which embody a novel approach to PDE-constrained optimization, dramatically reduce the limiting computational costs associated with studying large molecules.