July 06, 2011
GNS Healthcare is one of those companies that wouldn't have existed in the 20th century. It promotes itself as "a healthcare IT company that applies technology to optimize patient treatment." As such, GNS is at the forefront of a new era of drug development and delivery that is moving personalized medicine from theory into practice.
Headquartered in Cambridge, Massachusetts, GNS is the brainchild of Cornell physicists Colin Hill and Iya Khalil, who founded the original company in 2000, under the name Gene Network Sciences. Hill is now the CEO and president of GNS and Khalil is the company's Executive VP. Their idea was to exploit supercomputing technologies, in the form of "big data" analytics, to identify genetic biomarkers for drug efficacy.
Such an approach requires the ingestion of large volumes of genetic and clinical data, along with lots of data-intensive processing, both of which were expensive propositions a decade ago. In 2000, a modest-sized cluster with a few dozen processors would take a year to analyze a person's genetic profile.
But technology has caught up to GNS' aspirations. Thanks to cheaper DNA sequencing technologies to generate the raw data and much more powerful (and less expensive) high performance computing systems to process it, such analytics is now within the reach of commercial firms. Hill believes supercomputing, in particular, will enable advances in drug R&D that would otherwise have been impossible.
Much of that advancement is wrapped around the idea of personalized medicine. One of its principle tenets is to better match drugs to an individual's genetic makeup in order to make treatments safer and more effective. These compounds work at the molecular level and because even small genetic variations can produce big differences in a person's physical makeup, drug efficacy can vary significantly from one person to another. In a nutshell, the idea is to correlate these pharmaceuticals with a person's unique molecular characteristics.
Pharmaceutical companies, healthcare providers and patients all stand to benefit from better targeted drugs since, in theory at least, it drives down costs for everyone and delivers better results. Given the public's focus on reigning in healthcare expenses and the industry's concern with producing lawsuit-free drugs that will survive long enough to recoup development investments, a technology that delivers on both fronts would be welcome indeed.
To that end, GNS has developed a software platform that is able to analyze how genes, proteins and drugs interact in a virtual model. Dubbed Reverse Engineering/Forward Simulation (REFS), the software uses HPC clusters, or in some cases bona fide supercomputers, to sift through the data and figure out how all the bio-bits fit together.
In essence, the GNS software delivers a virtual clinical trial. But instead of taking millions of dollars and years to accomplish, the simulated version can be executed for a fraction of the cost in weeks or even just days. No one is ever put at risk, and there are no waivers to sign.
To accomplish this in silico, REFS creates a system interaction model of all the components represented by the data (reverse engineering) and then uses billions of queries (forward simulation) to reveal the most important genes and proteins driving those interactions. Importantly, it can also predict interactions for "what if" scenarios.
The technology was interesting enough to get the attention of DARPA, the US Department of Defense's research arm, which funded a case study on the GNS work. The effort, in collaboration with the Council on Competiveness, was part of a project to demonstrate the business case for high-end modeling and simulation technologies. This particular case study focused on a recent GNS collaboration with drug R&D specialist Biogen Idec.
The work with Biogen was to build a computational model for identifying novel drugs for rheumatoid arthritis sufferers. Today about a third of arthritis patients do not respond to the most commonly used anti-inflammation therapies (anti-TNF drugs). Since 1 to 2 percent of the world's population suffers from this condition, there is a lot of interest in developing more effective treatments.
The project with Biogen involved sifting through the genetic data from 70 arthritis patients to look for single nucleotide polymorphisms (SNPs), which are short sequences of DNA in which a nucleotide base in the sequence has been is altered. Gene expression data from the patients' blood as well as clinical information like pain levels, swollen joints, and other blood markers, were also encapsulated. Models were built from this data, which could then subsequently be used to conduct simulations with different drug compounds.
The data- and compute-intensive nature of the process is hard to fathom. Although only 70 patients were evaluated, it involved correlating hundreds of thousands of genetic variables on top of numerous clinical variables for each patient. Trillions of models were then constructed against each dataset. For example, REFS can simulate the "knock-down" of an individual gene by a certain drug, and then evaluate the result. With so many genes in the mix, the combinations can quickly escalate.
This was the first time a computer model of rheumatoid arthritis was developed that could be used to test new drugs and target pathways for individual patients. And it's not just that they've replace clinical trials with virtual ones. The sheer number of combinations that can be tested, not to mention the ability to virtualize risky drug scenarios means these simulations can go far beyond clinical testing. You just need enough computing horsepower make it work.
From Hill's perspective, the key technology to move this technology forward is high performance computing. "We have this strong conviction that the major game-changing advances in the biomedical sciences, drug development and patient care will not occur on a short time-scale without the extreme use of supercomputing," he says.
GNS itself has only a modest HPC setup, but its computational demands are nearly insatiable. Much of the time it uses big machines like IBM Blue Gene supercomputers (on-demand) and larger clusters from its partners. Besides Biogen Idec, Johnson & Johnson and Pfizer have teamed with GNS on other drug R&D projects, and the company is also engaged with a number of academic and non-profit research organizations.
If solutions like that from GNS deliver on their promise, they will have arrived in the nick of time. Skyrocketing labor and drug development costs and aging populations are straining healthcare delivery in much of the developed world. For less economically fortunate nations, 21st century healthcare is simply out of reach. For both rich and poor, the era of personalized medicine can't happen too soon enough.
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?