Lillian Chong Focuses on Underexplored Regions

Science and writing combine two of Chong’s passions. She works at the forefront of expanding the possibilities of dynamic simulations like those wriggling protein molecules. Recipient of numerous awards – including an NSF CAREER Award and Pitt’s Tina and David Bellet Teaching Excellence Award – Chong develops tools powerful enough to create biomolecular simulations that model the full compass of critical biological events.

“Science is creative,” Chong explains to the class. “Science and writing can reinforce each other.”

In the program she developed, while she was demonstrating a simulation of molecules students work on poems describing molecular level emotions.

Despite the existence of hardware capable of massive computation, Chong, a Pitt CRC collaborator who serves on the Center’s Advisory Committee, points out that hardware advances alone are not sufficient for computer simulations of biological systems. For example, simulations of relatively small proteins have required expensive, specialized supercomputers to reach timescales of just milliseconds. Many biological processes occur on much longer timescales. Simulating the emergence of rare events such as protein folding requires more than state-of-the-art hardware.

Chong is exploring another approach. Along with Daniel Zuckerman at Oregon Health & Science University, Chong led the development of an open-source interoperable software package with fancy algorithms for enhancing the efficiency of creating and analyzing simulations, known as WESTPA (Weighted Ensemble Simulation Toolkit with Parallelization and Analysis, with a nod to Western Pennsylvania). Chong and Zuckerman initially described the software in a 2015 publication in the Journal of Chemical Theory and Computation.

At the core of WESTPA is a weighted ensemble path sampling algorithm. Essentially, the weighted ensemble algorithm enhances modelling rare events at any level of detail ranging from atomic to cellular and beyond by focusing on sampling transitional pathways, rather than already known stable states. During simulations, existing known pathways are pruned away to focus computation power on underexplored regions. WESTPA focuses on the mechanisms of transitions, creating multiple simulations on multiple paths to reveal the most likely paths and the events affecting the rates and outcomes of transitions.

WESTPA is an attractive tool. It is primarily written in Python, an accessible yet powerful computer programming language, and can be used on any computing resource including typical academic clusters and supercomputers. It interfaces with any stochastic sampling engine. Chong, Zuckerman, and Jim Faeder (Pitt Computational and Systems Biology) have been awarded a four-year NIH RO1 grant for the continued development and maintenance of WESTPA, including enhancing its interoperability and scaling.

WESTPA can expand its capability across thousands of processors. To help accommodate that volume of computation, funds from the grant were provided to Pitt CRC to expand its graphics processing unit (GPU) nodes. Based on technology originally developed for video games, GPUs accelerate simulations by an order of magnitude beyond traditional computing that uses central processing units.

“Adding GPUs is a game changer,” Chong says. “With GPUs, we can run simulations on the microsecond timescale in one to two weeks. Coupled with WESTPA, the resulting weighted ensemble simulations can reach the seconds timescale, which is relevant to the action of drug molecules and has not been attainable using standard simulations on even the most sophisticated hardware.”

This article has been coppied from CRC News.

Originial article can be found here