YouTube Facebook LinkedIn Google+ Twitter Xinginstagram rss  

Protein-Powered Drug Discovery

Research & Discovery
Winner: Massachusetts Institute of Technology
Project: Identifying Drug Effects via Pathway Alterations using an Integer Linear Programming Optimization Formulation on Phosphoproteomic Data

By Alissa Poh

July 29, 2010 | Whole-genome methods have garnered a lot of attention in target-based drug discovery, leaving their protein-based counterparts somewhat squeezed in the shuffle. Nevertheless, one promising proteomics approach received the Best Practices Award for Research and Discovery.

This new tool for understanding how drugs work uses novel algorithms that deconvolute data derived from phosphoproteomics—a branch of large-scale proteomic studies that identifies and characterizes protein activity. Details were published in the December 2009 issue of PLoS Computational Biology.

Alexander Mitsos and Leonidas Alexopoulos, the paper’s lead and senior authors respectively, describe phosphoproteomic measurements as “the ultimate reporters of a drug’s action on signaling networks.” Changes in phosphorylation status almost always reflect changes in protein activity, which provides clues as to what molecular pathway(s) are activated and what proteins might be potential drug targets.

As a postdoc at MIT and Harvard Medical School, Alexopoulos and colleague Julio Saez-Rodriguez pursued ways to create cell signaling topologies, or pathway maps, using high-throughput phosphoproteomic data.

Shortly before returning to his home country, Greece, in 2008—where he is now a lecturer and group leader of the Systems Biology and Bioengineering Laboratory at the National Technical University of Athens—Alexopoulos struck up a conversation with Mitsos, then a research group leader at Germany’s RWTH Aachen University.

Mitsos had developed what Alexopoulos called “some neat computational approaches,” and they both saw an opportunity to use these to generate cell signaling topologies.

Despite their different scientific backgrounds, they found a good niche for collaboration. Alexopoulos pooled his training as a biological engineer with Mitsos’ computational knowledge. Together, they created a two-step system: first build pathways simulating cell function, then identify drug-induced alterations of those pathways.

Mitsos, Alexopoulos, Saez-Rodriguez, and other team members Ioannis Melas, Paraskeuas Siminelakis, and Aikaterini Chairakaki used their method to evaluate the effects of four drugs on HepG2, a liver cancer cell line. They picked 13 key interaction points, identified by phosphorylation of specific proteins. They then tracked these under different conditions, treating the cells with various combinations of cytokine ligands, kinase inhibitors, and four anti-cancer drugs: Lapatinib, Erlotinib, Gefitinib (all three block EGFR), and Sorafenib, a “dirty” inhibitor that blocks signaling via several kinases, including RAF.

The researchers were able to build a HepG2-specific map which confirmed known effects such as signaling responses to ligands, and the drugs’ main targets. They also uncovered an intriguing off-target effect of Gefitinib, observing that the drug inhibits signaling downstream of the protein cJUN along the interleukin-1 alpha pathway, possibly by interacting with molecules upstream of cJUN.

“This was pretty surprising,” Alexopoulos says. “Gefitinib is known as a pure EGFR inhibitor, and I wouldn’t have thought that it would also act along a totally different pathway.”

This strategy, Alexopoulos says, is “significantly different” from others currently implemented for drug screening—for instance, high-throughput in vitro assays—which generally identify binding partners for drugs of interest. It’s more of a physiologic approach. “We’re answering the same question from an alternate point of view; both are complementary methods,” he says.

The key strengths of this tool lie in its scalability and reproducibility; it allows precise, defined measurements to be made within a well-curated system. Besides being useful for validating known drug actions, it has the potential to identify drug effects missed by current screening procedures, as demonstrated by detection of Gefitinib’s off-target activity against cJUN.

So what was the biggest challenge they faced in developing this system? “Definitely language,” chuckles Mitsos, now an assistant professor of mechanical engineering at MIT. He’s a math person, while Alexopoulos focuses more on biological systems. “We talked a lot and tried to solve problems as they came, while always trying to not impose any particular method or point of view.”

This article also appeared in the July-August 2010 issue of Bio-IT World Magazine. Subscriptions are free for qualifying individuals. Apply today.


Click here to login and leave a comment.  


Add Comment

Text Only 2000 character limit

Page 1 of 1

For reprints and/or copyright permission, please contact Angela Parsons, 781.972.5467.