Clinical OMICS

JAN-FEB 2019

Healthcare magazine for research scientists, labs, pathologists, hospitals, cancer centers, physicians and biopharma companies providing news articles, expert interviews and videos about molecular diagnostics in precision medicine

Issue link:

Contents of this Issue


Page 18 of 51 January/February 2019 Clinical OMICs 17 [Since then] several other trends have kicked in that lead directly to translational medicine. As the cost of obtaining the sample began to drop signifi- cantly—and the ability to process data began to increase significantly—the value of the data began to rise rapidly. And that gets us into translational informatics. That is, to take information from the research phase and apply it directly to patient care or drug discovery. Why is translational informatics gaining prominence? Daly: A couple of factors. One, the cost of obtaining genomic data has dropped enormously, so there is much more data available. Until recently, there were very few large scale data sets that could provide both the geno- typic and phenotypic data. But, there are now as the cost has declined. Second, the compute power now avail- able in the cloud allows you to do large-scale analyses. Some of the customers are doing analysis using 10,000 or 20,000 servers, and there is no way you could ever do that locally. So you have both the decline in the cost of obtain- ing data, and the increased ability of the tool to use it. There is a growing awareness that for many core disease states—and the frontrunner is oncology—understand- ing a patient's genome, or the tumor genome, is foun- dational to figuring out how to develop drugs and treat these patients. Analysis that includes phenotypic data is crucial for precision medicine research. What is the source of your phenotypic data? Daly: The most important dataset we are using and processing right now comes from the UK Biobank. They have built a cohort of 500,00 individuals, who have donated some 20 million physical samples, that are being sequenced and interfaced with their patient records. So the ultimate source of the phenotypic data are the records obtained via the UK Biobank. Is machine learning already playing a role? Daly: Machine learning is already an important appli- cation for us. It turns out machine learning does a much better job than human-coded algorithms on reducing the output of the sequencer to an actual variant file. We had some influence and worked on that with Google for [their tool], which is called DeepVariant. I predict within two or three years, machine learning will replace most of what is currently used. What would you say are the three biggest benefits of translational informatics? Daly: Not in order of importance, I would list these: First, there is a very important role for this in figuring out how to stratify clinical trails. In most of the trials a substantial number of the patients are non-responders. I think now it is becoming clear to everybody that if you (continued on next page) Nipitpon Singad / EyeEm / Getty Images

Articles in this issue

Links on this page

Archives of this issue

view archives of Clinical OMICS - JAN-FEB 2019