Rice cluster an essential tool for professor’s drug discovery research
March 3, 2016
Chemical compounds that bind to a protein in treating one disease are likely to bind to similar proteins — and potentially to be useful for treating other diseases.
Purdue Professor Gaurav Chopra is using nature’s tendency to use basic components across species and purposes, a product of evolution, along with Purdue’s Community Cluster Program research supercomputers to discover new drugs and new ways of employing existing drugs. He’s already identified a well-tested cancer treatment as being effective at treating Type 1 diabetes.
“When a drug enters the human body or any organism it does not interact with just one or two proteins,” says Chopra, an assistant professor of physical chemistry. “It actually interacts with multiple proteins, which I refer to as the ‘interactome.’”
Chopra, one of the new faculty members involved in President Mitch Daniels Purdue Moves program, which includes a drug discovery initiative, uses computational chemistry and bioinformatics to build what he calls an “interactome signature” for chemical molecules.
The basic hypothesis is that if two compounds have similar signatures they will behave in a similar manner for therapeutic use. The interactome signature contains information about targets, and about off- and anti-targets, which can prompt undesirable side effects. This approach has been used to repurpose drugs, but the methodology and algorithms are general to discovering and designing novel drugs where there’s probability of finding a drug in a chemical dataset or the best lead compound fragments for a target are already known. Chopra’s computational work informs a wet lab component of his research, up to pre-clinical testing of treatment possibilities in mice.
“Everything in my lab, or almost everything, starts with a computational prediction to generate a hypothesis and then we go and validate those hypotheses,” Chopra says.
High-performance computing systems like Purdue’s Rice community cluster, named to the TOP500 list of the world’s most powerful supercomputers in 2015, are essential for his research.
“It did play a role in my decision to come to Purdue, in that, the things are set up in a way where I can come in and start the computational part of my lab, which I've already done,” says Chopra, who arrived in January from the University of California at San Francisco.
In one instance, for example, Chopra modeled 4,000 human ingestible compounds and a battery of 50,000 proteins each with five binding sites on average, making for a billion interactions that had to be computed even before integrating a substantial amount of disease data.
“It requires a huge computational resource to do this at a large scale where we can get this systems view,” Chopra says. “It’s a very engineering approach to drug discovery in some ways but it’s strongly rooted in science.”
He uses a variety of codes developed by his lab and others for protein structure prediction, modeling interactions and data integration, all tied together by machine learning to build a pipeline to results.
More typically in drug discovery, compounds are developed and screened for interaction with a single protein, but 95 percent of the time the compound selected by the screening fails as a drug for reasons including being ineffective or having undesirable side effects. Probing lots of possibilities at once, as Chopra does, ups the odds of discovering a new treatment.
When a compound has been approved previously for one use, as in the case of the cancer drug Chopra found to be effective against Type 1 diabetes, it also can shorten the time to regulatory clearance for other uses, because the answers to questions like how toxic the chemical is to humans already are well understood.
Chopra also plans to apply his interactome approach to cell-guided therapy, an emerging frontier in medical treatment. The idea is to unite therapeutic compounds and cells, which would act as natural biosensors to carry the drugs to disease targets. However, it is important to understand the litany of interactions those cells have with non-targeted aspects of the body and use this information to make design choices in order to avoid unintended consequences.
For information about Rice, the big memory Snyder cluster designed for life science research, and Purdue’s other Community Cluster Program research supercomputers, email email@example.com or contact Preston Smith, director of research services and support for ITaP Research Computing, 49-49729 or firstname.lastname@example.org.