Supercomputing poised for leading R&D role in personalized medicine

Humans just aren't smart enough to realize the full potential of personalized medicine alone. And while modern sequencing platforms yield huge amounts of molecular data to advance studies, researchers have a long way to go before many of those data can be translated into cures. Why not inject a greater dose of supercomputing and machine learning into the equation?

Here's why: Our collective understanding of human biology represents a tiny percentage of the mechanisms and networks that are actually at play in diseases. What drugmakers and medical researchers are aiming for nowadays is personalized medicine, a way of treating diseases that calls for a much deeper understanding of human biology than exists today. And the industry seems to know that past approaches to R&D aren't going to expedite our journey toward mainstream availability of personalized treatments.

The good news is that there are at least several reasons to believe that R&D groups are going to move faster toward personalized medicine than before. For one, DNA sequencing and other molecular research tools have become cheaper and faster. What's more, there's pressure from governments, healthcare payers and patients on drugmakers to deliver new medicines with greater safety and efficacy than existing treatments. Also now available are smarter computer systems, which are getting better at helping researchers deal with the complexity of the human genome and other types of data at our fingertips to fully realize the promise of personalized medicine.

Colin Hill, chief executive and co-founder of GNS Healthcare, has been working for years, with some success, to convince drugmakers to give his Cambridge, MA-based firm's supercomputing technology a shot. "The computational technology is finally there to crunch this data and really determine what the causes of certain disease outcomes are, which gives us the biomarkers of drug response and drug targets that can make the next-gen drugs much more safe and efficacious than anything we have now."

As I learned during a recent visit to GNS, the firm's reverse engineering forward simulation technology doesn't rely on existing scientific knowledge about human biology to provide insights about circuitry driving disease and how well a drug works for a given patient. Rather, the firm's supercomputing-enabled REFS system can simultaneously crunch multiple layers of raw genotypic and molecular profiling and clinical outcome data to yield detailed disease models. With Biogen Idec ($BIIB), for example, the company reported in March that this approach helped uncover novel drug targets and markers for the approximately one-third of rheumatoid arthritis patients who don't respond to standard anti-TNF drugs.

To be clear, Weston, MA-based Biogen isn't the only drug developer that has given supercomputing a chance. GNS has also worked with researchers at such firms as Johnson & Johnson ($JNJ), Novartis ($NVS) and Pfizer ($PFE). And over the past week, we've learned that London-based drugmaker AstraZeneca ($AZN) plans to establish a new R&D center in St. Petersburg, Russia, to develop data analysis and bioinformatics tools to predict drug safety and efficacy. These companies spend billions on R&D, and they are looking for ways to use advances in computing to help them be more productive in their wet lab experiments and ultimately gain better odds of success in clinical trials.

"Computational engines like what we have work. It's now been validated in enough systems," Hill says. "Does it need more validation? Yes. Are we in the process dong that? Yes. But such a platform didn't exist 5 years ago, 10 years ago. This stuff just couldn't be done before."

Make no mistake: I don't think GNS's technology is going to make drug development easy or translational research simple. But there's a big opportunity here to make machine learning and supercomputing bigger parts of making personalized medicine a reality. - Ryan McBride (twitter | email)

P.S. FierceBiotech IT in Chicago for DIA 2011 - This week Chicago hosts DIA's 47th Annual Meeting, and the FierceBiotech IT crew is in the Windy City to be a part of it. George Miller is leading our coverage so drop him a note if there's something in particular we should check out. In addition to George's coverage, I'll be tweeting occasionally (click to follow me) from McCormick Place. You can also browse all #DIA2011 tweets and follow all the action in real-time. And if you're at DIA 2011, we hope to see you this week! - Arsalan Arif, publisher (twitter | email)