Genomics research may transform personalized medicine. How can radiologists help?
Compiling samples of thousands of genomes from cancer patients may sound like a daunting task, but that's exactly what The Cancer Genome Atlas (TCGA) has been doing.
It's cataloging the DNA of tissue from thousands of patients, who suffer individually from 20 different types of cancer, ranging from glioblastoma multiforme to breast lobular carcinoma. Through this research, TCGA researchers are on a mission to answer one question with massive implications: What turns a normal cell into a cancer cell?
TCGA, which falls under the auspices of the National Cancer Institute (NCI) and National Human Genome Research Institute (NHGRI), is one of the major players in cancer genomics research. Its pilot program began in 2006 to characterize three heterogeneous tumors in the ovary, lung, and brain. In 2009, its mission expanded to encompass 20 different classes of cancer, each of which requires more than 500 tissue samples to prove significant findings.
Investigating cancer patterns deserves the attention of many organizations and diverse specialists in oncology, genetics, pathology, informatics, and even radiology. That's why this kind of research has spawned a new term: omics. It was named to encompass much more then genomics' traditional definition and scope. In that sense, omics incorporates the broad intersecting fields needed to study the human genome. If successful, this multidisciplinary, multi-institutional effort could revolutionize individualized cancer care and therapy by predicting what treatments and therapies may be successful. Yet, how can the radiology community help unlock the mysteries of cancer?
Describing Brain Tumors
TCGA's original scope did not focus on medical imaging, but it didn't take long for researchers to realize the importance of radiology in cancer patterns. After all, radiologic images provide another way of looking at it, and learning from, the mutations that cause cancer. Neuroradiologists like Adam E. Flanders, M.D., professor of radiology and rehabilitation medicine at Thomas Jefferson University in Philadelphia, are already marrying the imaging component to existing brain cancer genomics research data. "We're interested in why these malignant brain tumors can appear so radically different in MRI scans in different patients," he says, adding that glioblastomas are typically classified in a single general category by neuropathologists, even though we know the disease has many variations that directly reflect their variable genetic structure. "In much the same way that our genetic structure, or genotype, defines how we look, from the shape of our fingers to the color of our eyes, the genetics of cancer determines the appearance or behavior of a tumor like glioblastoma."
Flanders and a team of neuroradiologists with backgrounds in oncologic imaging were assembled from various institutions around the country, such as the National Institutes of Health, University of Virginia, Emory University, Henry Ford Hospital, Brigham & Women's Hospital, and Stanford University, with the goal to actively study brain tumor images and create a controlled vocabulary for describing their phenotypic appearance on an MRI. Typically, a phenotype is viewed through the genetic lens, resulting in visible traits like blue or green eye color. But a "phenotype is also the way a disease expresses itself," explains C. Carl Jaffe, M.D., professor of radiology at Boston University School of Medicine and consultant to the NCI Cancer Imaging Program (CIP). A brain tumor phenotype is simply those features that a radiologist actually sees in the image — or in Flander's case, a brain MRI. To assess glioblastomas in brain scans more efficiently, for example, imagers need to use the same set terms to describe what they see. "We need a very controlled, semiquantitative method to describe the heterogeneous appearance of these tumors," says Flanders. For example, what actually defines necrosis? Necrosis requires a standard definition as does each way a radiologist might report tumor appearance.
This insight sparked a mission for Flanders and his colleagues: to come up with a comprehensive list that describes each element of how a brain tumor appears on an MRI. Then, Flanders and the coinvestigators could look at each case, apply the developed lexicon, and score each glioblastoma in a consistent manner. This strategy has already achieved a high level of agreement among neuroradiologists involved in assessing TCGA's Internet-accessible repository of brain tumor images, many of which are publicly accessible at http://cancerimagingarchive.net.
Connecting Imaging to Genetics
Flander's research is the first of its kind and an exemplary illustration of how radiologists can contribute to omics research. But how can his studies lead to more personalized cancer care? First and foremost, such improvements in care require collaboration between a number of specialties, including radiology and pathology. According to Joel H. Saltz, M.D., Ph.D., professor and chair of the department of biomedical informatics and professor of pathology at Emory University in Atlanta, who is also involved with genomics research through TCGA and the CIP, "Understanding the heterogeneity of a tumor is very important. And pathologists can contribute [to the cause] because of their expertise in doing image analyses of individual cells in tumors," he explains. "But if you have a large enough region, you can look at the relationship between changes you might see in a diagnostic imaging study and relate it to the subclasses defined from the geneticists' point of view." In other words, radiology provides a broader view of a tumor, which Saltz compares to the view from a satellite. Studying the pathology or genetics of a tumor, however, is more lie the view from an airplane.
Flanders elaborates, "Imaging is complementary to genetics. [With both], we can get a much better handle on how patients might respond to a particular treatment or survive." Researchers hope one day to achieve enough evidence-based support to say, for example, "If a tumor looked this way in an MRI, the pathology looked this way, and these gene clusters were turned on, then a patient will do better with treatment A versus B," emphasizes Flanders.
Not only are radiologists, pathologists, and geneticists involved in these efforts; these researchers also rely on informatics and statistics experts to look for significant relationships between the tumors studied. Flanders and Saltz also work with the Cancer Biomedical Informatics Grid® (caBIG®) — a virtual network of data, individuals, and organizations redefining how cross-disciplinary cancer research is conducted. "We need to create tools to bring all resources available in cancer research closer to patients," Flanders explains. "And we need to work with pathologists, informaticians, and biostaticians," he adds, because these specialists can tell us whether research findings make sense and if a scientifically robust conclusion is generated.
It's important to acknowledge that true scientific conclusions about cancer genomics will take years, probably decades, to discover. However, that doesn't mean that current omics research isn't advancing quickly. According to a 2008 article in Nature, "... if researchers are surprised at how quickly genome-wide association studies have become consumer products, they need to realize that things will only move faster in the future, with findings moving from the lab to Internet chat rooms and to people's lives with astonishing rapidity."1
Genomics certainly is being brought into the public eye and has even been featured in the ever popular TED — Technology, Entertainment, and Design — conference series. In a lecture given as part of TED talks, Richard Resnick, CEO of software company GenomeQuest, gives two examples of patient diagnoses made through genetic profiling. The fact that these advances are already altering patient care and personalized medicine is a great indicator of what's to come. And, as Jaffe explains, "Imaging is a noninvasive way to characterize the human body." The fact that imaging is also often a part of patient workups allows for enormous contributions from the radiology community.
Realizing this potential, the ACR convened a meeting of leaders in October 2011 to discuss genomics research and the role of imaging experts in the omics field. The group determined several action items to strengthen radiology's role in collaborative research with the NCI's TCGA, CIP, and caBIG, including sharing current research, developing team coinvestigators, publishing early findings in medical literature, and encouraging researchers to contribute to, use, and cite images from CIP's archive. Most importantly, radiologists need to "keep their antennas up," when it comes to this field, says Jaffe. And as Resnick concludes, "I urge you all to wake up and to tune in and to influence the genomic revolution that's happening all around."
1. “Getting Personal,” Nature 2008;455(7216):1007.
By Alyssa Martino