Sometimes great discoveries come from not knowing where to look.
ISSAC Corp., a Springs-based advanced data analytics and system-of-systems engineering firm, recently made its first foray into big data analysis for the medical world — and ended up in the famed British Journal of Cancer.
ISSAC’s analysis revealed important new findings in how pancreatic cancer patients react to clinical trials. The work was part of a collaborative data analytics effort with the Mayo Clinic and the Translational Genomic Research Institute.
The results of the research, “Clinical study of genomic drivers in pancreatic ductal adenocarcinoma,” were published in the British Journal of Cancer last month. ISSAC chief scientific officer Ray Deiotte is listed as the study’s second author, and ISSAC appears after the Mayo Clinic in the list of institutions.
“The beauty about the way we analyze data is we look at information a little differently than the traditional processes for advanced data analytics — we really focus on letting all the data tell us the story. We do that through artificial intelligence and machine reasoning capabilities we have built into our platform,” said ISSAC CEO Tim Jones, who is also credited as an author on the paper.
Over two years, ISSAC spent about 350 hours analyzing data from two historical pancreatic cancer clinical trials using VOR, its end-to-end analytics solution named after the Norse goddess of wisdom and knowledge.
“Out of this emerged some interesting data relationships that we couldn’t interpret because we’re not bioinformaticists or biologists, so as we were identifying relationships and patterns within the data, we were feeding those back to Dr. [Daniel] Von Hoff [of TGen] and Dr. Michael Barrett at Mayo,” Jones said.
The data had previously been analyzed using more traditional data analytics techniques, but ISSAC’s analysis uncovered new causalities, correlations and identifiers that hadn’t been seen before.
“The very initial response [from researchers] was, ‘Why are you looking at that piece of the data? We don’t believe there’s anything there,’” Jones recalled.
“And the answer from us, ‘We don’t know any better.’
“One of the things that we truly believe is that human bias can negatively affect the way data is analyzed, because we’re experts at what we’re experts at. By having that bias, we automatically disregard data points because of what we’ve seen and what we’ve known for the last 20-30 years.”
Jones said with VOR, ISSAC is able to remove human bias from the analysis process to avoid influencing or disregarding information contained in big data.
Merriam-Webster defines big data as “an accumulation of data that is too large and complex for processing by traditional database management tools.”
Michael Ames puts it this way: “When your computational problem exceeds your computational capacity to solve it, then you have a big data problem.”
Ames is associate director of Health Data Compass, an enterprise health data warehouse headquartered at the University of Colorado Anschutz Medical Campus and jointly sponsored by University of Colorado School of Medicine, CU Medicine, UCHealth and Children’s Hospital Colorado.
“We’re at the point,” he said, “where the amount of medical data available is 10 or 100 or 1,000 times beyond … the ability of a doctor to comprehend, and beyond the ability of a researcher to look at and draw reasonable conclusions” without turning to technology to unearth hidden patterns and correlations.
“So we’re applying big data technologies against all of this data with the goal of finding out, in this massive volume of data, or in this whirlwind of high velocity data: What is the thing that matters?” Ames said.
Removing human bias is part of an important transition in how big data is analyzed, Ames said.
“What is changing is a dramatic and rapid embracing of machine-learning techniques in order to develop predictive models that are proving, in many cases, to be more accurate than what the experts might have hypothesized on their own — and can often reveal unexpected patterns and relationships in the data,” he said.
Machine learning feeds the data into advanced computational algorithms, en masse, without giving the computer a hypothesis.
“Through massive and iterative computational processes, without that hypothesis in place, [the machine learning system] goes and finds those relationships … within the data that we would never have been likely to imagine,” Ames said.
In the case of the pancreatic cancer study, VOR revealed new correlations within genomic data, as well as among other human factors such as gender, height, weight and lifestyle, which the researchers believe affect how patients reacted to clinical trials.
Based on these discoveries, the Mayo Clinic and TGen are already identifying new avenues for clinical trials and possible novel treatments for pancreatic cancer patients.
Considered particularly aggressive, pancreatic cancer is the fourth leading cause of cancer deaths in the United States. Patients often die shortly after diagnosis, because the cancer usually isn’t discovered until it has metastasized.
The success of this study is steering ISSAC toward more work with health organizations, including new projects with the Mayo Clinic, the Gates Foundation and TGen.
Founded in 2007, ISSAC started with Department of Defense work and has evolved over the years to work in energy, intelligence and cybersecurity.
“We’ve always known our analytics tool could have significant impact in the medical world and health world,” Jones said. “We knew there were avenues there … but our first opportunity was this effort with Mayo and TGen. We’re very excited about it. We’re passionate about continuing down that road.”
Collaboration among organizations is critical in successful big data analysis for medicine.
“It’s necessary,” Ames said. “Any one hospital’s view of a patient is a slice of Swiss cheese because people only go to hospitals at certain times for certain conditions.”
A patient’s total care is likely shared between multiple hospital systems, a system of primary care providers and local clinics, as well as research studies and online pharmacies.
“To get a complete picture of an individual’s health care requires the integration of data from across many, many sources,” Ames said. “And then to gather enough data about individuals to be able to draw meaningful conclusions about how we can improve the health of populations as a whole, requires the contribution of many, many individuals’ data.
“So the objectives of big data analytics and health care literally cannot be accomplished without the collaboration of multiple institutions locally, regionally, nationally — and that is absolutely happening in Colorado.”
The work at the University of Colorado Anschutz Medical Campus embraces the idea that “the answers will be in the data,” Ames said.
“This institution, along with our partner hospitals in UCHealth and Children’s Hospital Colorado, have embraced this notion and are investing heavily to drive breakthrough medical advances in this space, by applying great technology and science to this massive onslaught of data,” he said.
“As we speak to our colleagues across the state we’re all beginning to speak the same language and coalescing around strategies for the use of big data to drive breakthrough medical advances.
“So this is a culture where this kind of work is happening, where innovation is occurring, where the job market is ripe for people who have these skills.”