1. DiGiulio, Sarah

Article Content

Is genomic analysis a better way to categorize a tumor than its tissue of origin? For some patients, yes-according to new research that proposes a new way to classify cancer that could more accurately predict clinical outcomes for certain subsets of patients.

Chuck Perou, PhD. Ch... - Click to enlarge in new windowChuck Perou, PhD. Chuck Perou, PhD.

In the study (Cell 2014;158:929-944), a total of 3,527 cancer specimens from The Cancer Genome Atlas that came from 12 tumor types were classified by researchers into 11 major subtypes based on their molecular characteristics. And for approximately 10 percent of those patients, the clinical outcome predicted by the molecular reclassification was significantly more accurate than the clinical outcome predicted by the original tissue-of-origin classification-even when considering all the previous clinical knowledge.


"Ten to 20 percent of patients may be having their tumors misclassified and hence may not be getting the optimal treatment," study coauthor Chuck Perou, PhD, Professor of Genetics and Pathology and University of North Carolina School of Medicine Breast Cancer Program Leader, both at UNC Lineberger Comprehensive Cancer Center, explained in a telephone interview.


Five genome-wide platforms and one proteomic platform were used to analyze the specimens: whole-exome DNA sequence, DNA copy-number variation, DNA methylation, genome-wide mRNA levels, microRNA levels, and protein levels for 131 proteins and/or phosphorylated proteins. And to investigate the clinical relevance of the new subtypes, a Kaplan-Meier Survival analysis was performed on the samples and showed that both the conventional tissue-of-origin and new genomic categorizations were both prognostic and provided independent information.


The key findings were that:


* The integrated multiplatform analysis of the samples provided independent and clinically relevant prognostic information above and beyond tumor stage and primary tissue-of-origin; and


* One in 10 cancer patients would be classified differently by this new molecular taxonomy versus the current tissue-of-origin tumor classification, and if used to guide therapeutic decisions, this reclassification would affect a significant number of patients to be considered for nonstandard treatment regimens.



The research by Perou and his colleagues was selected earlier this year for inclusion in the American Society of Clinical Oncology's Clinical Cancer Advances 2015 report, which documents important progress being made in clinical cancer research and highlights emerging trends in the field (OT 2/25/15 issue).


More on how the researchers analyzed the data and determined the new classification system, as well as how the 12 conventional tumor types fit into those new subtypes is detailed in the paper. Meanwhile, we asked Perou about how these findings can and will affect clinical practice-and if such a system is ready for prime time.


1. Can you elaborate on why this research is significant? How do the findings relate to drug development, cancer diagnosis, and prevention?

"It's going to help all of those things. This is a valuable method of classification-and potentially very reproducible and accurate. It is a first and critical step toward personalized medicine.


"By using these many technologies, we now know the common genetic mutations that occur in each of these disease groups. We can try to partner each of these disease groups with the drugs that target these common mutations and the common pathways that are altered within them.


"This [research] holds the promise of being the new foundation of personalized medicine that not only includes just DNA sequencing, but also utilizes gene expression and other means of looking at the DNA as well-providing a link between the many different available technologies, and more importantly, many different parts of the tumor.


"A prime example is the immune system. There have been some extremely exciting advances in the treatment of melanoma by targeting the immune system, not the tumor cells. And as part of our classification you could also see that certain of these molecularly defined tumor subtypes had close correlations with the presence or absence of immune cells.


"This [reclassification] gives us a more complete picture of each of these types of cancer. And that's really going to be important for optimal therapeutic targeting."


2. Is this new molecular taxonomy ready for prime time?

"Whenever you discover something in the laboratory it takes a few years to get it to the clinic because it needs further validation and further testing. Some of the differences in classification between sites of tumor origin versus the molecular classification were known previously and are being tested prospectively in clinical trials-so, on that level we're already doing some of this clinical validation work.


"There are some barriers, including cost, because we're using five technologies [to analyze each cancer cell] instead of one. But costs are coming down and I think-and hope-that in a few years we will more often be using these types of 'multi-analytic' molecular tests because they are improving patient care by finding these more homogenous, biologically related disease groups.


"This paper really sort of puts many individual studies together into one big study, and I do think it suggests we should be re-evaluating cancer patients' classification based on this new scheme-and we're going to try to make it happen."


3. What's the next step?

"There's two ways to do this more rigorous validation research. One is to actually do prospective clinical trials where you test the new biomarker versus the standard of care to see if it improves patient outcomes. That's expensive, though, because you have to run brand new clinical trials.


"A more intermediate approach is to retrospectively analyze existing clinical studies-if those studies collected tumor specimens. Many of those patients' outcomes are already known. So you can apply your patient predictor in a predetermined manner-i.e. you can pretend the study was run in a prospective manner and you get a result much quicker.


"Many of these genomic assays are currently being tested in this retrospective manner, and there are also some prospective studies being planned for some of the DNA-based markers."


More from OT Associate Editor Sarah DiGiulio's '3 Questions on...' blog can be accessed at