Matchmaking for Science
RENCI connects researchers with computing resources on the Open Science Grid
 |
A De Novo designed protein-protein |
Computing enables much scientific research today, but finding the right high-end computing resources is often a challenge.
Brian Kuhlman, a biochemistry and biophysics professor at the University of North Carolina at Chapel Hill and head of the Kuhlman Laboratory in the UNC School of Medicine, needs thousands of compute hours each year to create modules of different protein configurations and determine their abilities to combine with one another. It’s work that could lead to new treatments, and eventually cures, for serious diseases such as diabetes, Alzheimer's, HIV/AIDS and many cancers. And it requires computing power well beyond that of his own small research lab.
Catherine Blake, an assistant professor in UNC's School of Information and Library Science, works with RENCI through the Faculty Fellows program to develop a method for retrieving, analyzing and finding relationships in published research texts from multiple disciplines. The work could help researchers find research results in other fields that relates to their own fields of study and lead to more productive research practices and more cross-disciplinary sharing and collaboration.
 |
| Catherine Blake |
The project, called "Claim Jumping Through Scientific Literature" required a staggering amount of computing time in its first phase. With help from RENCI, Blake's team analyzed 162,000 documents from a variety of scientific fields using a process called dependency parsing, which analyzes the grammatical structure of sentences to find meaningful relationships among words in different documents and from different fields. Generating these results on a high-end desktop computer would have required about 15,000 hours--about 7 ½ years of 40-hour work weeks, said Blake.
Both Kuhlman and Blake looked to RENCI not only for the technical expertise of its staff and on-site resources, but also for its connections to a nationwide network of expertise and resources in computing and technology. Through RENCI, they found the Open Science Grid (OSG), a consortium of universities, national laboratories, scientific collaborations and software developers dedicated to meeting the ever-growing computing and data management requirements of scientific researchers. Supported by the U.S. Department of Energy Office of Science and the National Science Foundation, OSG provides access to its members’ independently owned and managed resources through a common grid infrastructure that uses high-performance networks to connect computing systems scattered across the country. As the leader of engagement activities for OSG, RENCI works with researchers to introduce them to the OSG and its resources and help them develop the skills needed to use the OSG national cyberinfrastructure.
 |
| Computationally designed protein that switches structure depending on environment. Click for larger image |
After being introduced to the OSG through RENCI, Kuhlman’s team was soon running large-scale jobs on the OSG with Rosetta, molecular modeling software used to study protein design, protein folding, and protein-protein interactions. Kuhlman's team used more than 150,000 CPU hours on the OSG in 2007—work that would’ve taken years on computers in the Kuhlman Lab.
For Blake and her research team, more than seven years of work was completed in about 28 hours using the grid resources of the OSG.
"We had the monumental task of parsing 162,000 documents with several thousand sentences each," she said. "It was an ideal project for the grid because the job is already carved up into small pieces that can use computer resources anywhere."
“Having RENCI here at Carolina helped us access the OSG," added Kuhlman. "That has been a huge time saver, but even more important, it has made it possible for us to examine questions that would otherwise be unanswerable,” said Kuhlman.
More information: http://www.renci.org/news/comp_match.php
~ Karen Green, Renaissance Computing Institute
March, 2008
This article also appeared in iSGTW.
Supported by the National Science Foundation and the U.S. Department of Energy's Office of Science