Medicine and Cognitive Computing

Stephen DeAngelis

January 14, 2014

Last fall an article in The Hindu reported, “Computer programs and algorithms developed in the last three decades are making realistic molecular simulations possible, which, in turn, are used to make drugs to cure chronic diseases such as cancer.” [“Computing to find cures,” by Raghavendra S., 15 October 2013] To underscore the point that computing has assumed a prominent place in the pharmaceutical field, the article pointed to the 2013 Nobel Prize for chemistry which was won by scientists who used computers (rather than real chemicals) to research chemical reactions. In the future, cognitive computing will be used to help create new medicines faster; but, perhaps the biggest medical breakthroughs will occur when data about medicines are merged with data about genetics to create truly personalized healthcare. Margaret A. Hamburg, a medical doctor and Commissioner of the Food and Drug Administration (FDA), asserts, “The difference between science and science fiction is a line that seems ever harder to distinguish, thanks in part to a host of astonishing advances in medical science that are helping to create a new age of promise and possibility for patients.”

Phys.org article states, “With the promise of personalized and customized medicine, one extremely important tool for its success is the knowledge of a person’s unique genetic profile.” [“New computing model could lead to quicker advancements in medical research,” provided by Virginia Tech, 4 November 2013] One of the first areas receiving a lot of attention (and money) is cancer research. Bruce Friedman reports, “Genomic sciences and the growing interest in precision medicine are having a variety of different effects on healthcare delivery. For example, patients are now being triaged into special clinics focusing on the most aggressive tumor types or tumors of unknown primary.” [“New Disease Targets for Old Drugs; Another Big Data Initiative,” Lab Soft News, 21 November 2013] Hamburg adds, “Today cancer drugs are increasingly twinned with a diagnostic device that can determine whether a patient will respond to the drug based on their tumor’s genetic characteristics.” She continues:

“Medical imaging can be used to identify the best implantable device to treat a specific patient with clogged coronary arteries; and progress in regenerative medicine and stem cell therapy using a patient’s own cells could lead to the replacement or regeneration of their missing or damaged tissues. Given these trends, the future of medicine is rapidly approaching the promising level of care and cure once imagined by Hollywood in futuristic dramas like Star Trek. But these examples are not science fiction. They are very real achievements that demonstrate the era of ‘personalized medicine’ where advances in the science of drug development, the study of genes and their functions, the availability of increasingly powerful computers and other technologies, combined with our greater understanding of the complexity of disease, makes it possible to tailor treatments to the needs of an individual patient. We now know that patients with similar symptoms may have different diseases with different causes. Individual patients who may appear to have the same disease may respond differently (or not at all) to treatments of that disease.”

Those types of insights and relationships are much more likely to be discovered by a cognitive computing system than by a physician. That’s why healthcare researchers and providers are welcoming new technologies with open arms. Although the concept of personalized medicines is exciting (imagine medicines with fewer side effects), cost remains an issue. Many people fear that personalized medicine will be available only for the wealthy. That may currently be true; but, at least costs in some areas are heading in the right direction. Take, for example, the cost sequencing the human genome. The Phys.org article notes, “Sequencing a genome … has gone from costing $95,000,000 to a mere $5,700.” I’m not sure I would have used the modifier “mere.” That’s still a lot of money for most people. Nevertheless, the drop in cost (and increase in speed) of sequencing has been amazing. The article goes on to state, “Now the research problem is no longer how to collect this information, but how to compute and analyze it.” Wu Feng, a researcher in the Department of Computer Science in the College of Engineering at Virginia Tech, provides an idea of how much DNA information is currently being gathered. He states, “Overall, DNA sequencers in the life sciences are able to generate a terabyte — or one trillion bytes — of data a minute. This accumulation means the size of DNA sequence databases will increase 10-fold every 18 months.” That is a lot of data. Only cognitive computing systems can garner insights and connections from that much data. The larger the database gets, the more precise the insights will be. Sophie Curtis reports, “Data-driven medical research is a growing field.” [“Data-driven medicine: understanding the link between genetics and disease,” The Telegraph, 16 October 2013] She continues:

“The Wellcome Sanger Institute, which was the single largest contributor to the Human Genome Project, is now using so-called ‘big data’ to investigate the genetic make-up of some of the most common causes of premature death. As one of the top five scientific institutions in the world specialising in DNA sequencing, Sanger Institute embraces the latest technologies to research the genetic basis of global health problems, including cancer, malaria, diabetes, obesity and infectious diseases. The hope is to one day understand the link between disease and genetics. The sequencing machines that run today produce a million times more data than the machine used in the Human Genome Project, and the Sanger Institute produces more sequences in one hour than it did in its first 10 years. … The data is analysed using the Sanger Institute’s supercomputer, which has 17,000 Intel processors and 22 petabytes of storage from DDN and other vendors. This data is then shared within the research community and through the Sanger Institute’s website, which gets 20 million hits and 12 million impressions each week.”

Tim Cutts, acting head of scientific computing at the Wellcome Trust Sanger Institute, offered two examples to Curtis showing how cognitive computing can advance medical treatments. The first example involved Addenbrooke’s teaching hospital in Cambridge.” He told her:

“Addenbrooke’s … had a few cases of MRSA in its neo-natal ward. Samples of the bacteria were sent to the Sanger Institute, which sequenced them in real time and identified that they were the same strain as each other, and that therefore there was a single source of the infection. Addenbrooke’s was then able to go through its records and identify where the bacteria had come from, meaning that the problem could be cleaned up and sorted out much faster than the hospital would have otherwise been able to do.”

The second example was even more telling. Cutts told Curtis about the “Sanger Institute’s discovery of a variant in a gene called BRAF, which was identified as being authored in 60 per cent of cases of malignant melanoma. This has led to the development of a treatment in just nine years, proving that rapid innovation is achievable with this kind of technology.” Cutts concluded, “The sort of statistical correlation analysis that we’re doing these days is an absolute classic ‘big data’ problem.” The Lab Soft News article highlights another use of cognitive computing in medicine: discovering new disease targets for old drugs. The report notes that such systems can utilize “some of the newly discovered genetic signatures for diseases and then harnesses the power of Big Data to match old drugs for new treatments of these diseases ” The future of pharmaceutical development is likely to look much different than it did in the past. Hopefully, this means that effective drugs will be able to be developed faster, with fewer side effects, at a reasonable cost.