New “big data” analytics company Via Science will market technology that allows drug sponsors to use the massive amount of complex data generated in research when making decisions about their drug development candidates and strategies.

The Cambridge, Mass.-based company, started by Gene Network Sciences co-founder Colin Hill, uses a supercomputer-based scientific platform called REFS (Reverse Engineering/Forward Simulation) to take large amounts of data and create predictive models that allow pharmaceutical and biotechnology companies to better match drugs to patients.

“High-powered machine-learning technology is being used in many, many places, from the routing of planes traveling around the world to the matching of ads to people searching on the Internet,” said Hill, CEO of Via Science. “Now it can be turned onto important and interesting problems such as ultimately improving patient outcomes.”

The REFS platform was developed by Gene Network Sciences, which is now a subsidiary of Via Science under the new name GNS Healthcare. Created to help streamline and accelerate the development of new drugs, the technology can help sponsors match drugs in development to patients who will respond by identifying the factors of predictive response. “Many drugs fail, not because they are bad, but because they are being given to the wrong people,” Hill said. “In the last decade, we have started seeing the beginnings of a more rational matching of drugs to patients, which ends up being key both in clinical trials and in patient care and reimbursement.”

REFS has been used to identify biomarkers that can predict how patients will respond to certain rheumatoid arthritis drugs for Weston, Mass.-based Biogen Idec. It also has been used for drug discovery and development by Pfizer and in oncology by Johnson & Johnson.

The ability to collect genetic and expression data, along with information from proteomics, imaging and clinical research, has given researchers greater insight. At the same time, the huge amount of data has created a challenge for drug sponsors. Traditional methods of data analysis, such as relying on the knowledge and intuition of experts to decide which molecules are worth measuring, or using basic statistics and bioinformatics—early forms of machine-learning—to identify molecules that seem to line up with the outcomes, are insufficient to process the volume and complexity of the drug research data generated today. “If it was just a single layer of data, like DNA sequence variation together with endpoints, one can get by with using more simple bioinformatics techniques. But when you have multiple layers of data, it becomes very complicated,” Hill said. “The world is different than it was five years ago. We have much, much more data available to us now to make clinical trials smarter.”

The Via Science technology, which uses so-called “big data” tools and processes, extracts cause-and-effect relationships from data and uses an unbiased, data-driven approach for its analysis. “We want to get at the underlying circuitry of the system, similar to what a human would try to do based on expert knowledge, but we want to do that in an unbiased way,” said Hill. “We can, and sometimes do, use prior knowledge in our discovery. But we take an approach to discovering biological knowledge and particular biomarkers and treatment algorithms—sometimes even drug targets—that starts with the data and allows the answers to bubble up from sets of well-characterized clinical, pre-clinical and patient-care data sets.”

Via Science will serve as part holding company for Hill’s previous two companies, Fina Technologies, which applies the technology to financial trading, and GNS Healthcare. Hill also will use the company as part-incubator for launching new startups related to supply-chain management, insurance risk assessment and consumer marketing.

—Karyn Korieth