February 1, 2014 (Vol. 34, No. 3)

MaryAnn Labant

Vast amounts of genomic discovery research have yet to be translated into routine clinical use. In the meantime, scientists continue to advance molecular diagnostics, building evidence to support its use in predicting the best therapeutic approaches.

Each tumor is a unique evolutionary process with an individual genealogy. Comparing the DNA sequences of tumors with those of normal tissues typically gives little information about disease-process dynamics. RNA profiling can be enlightening, but this technique, like DNA sequencing, is complicated by enormous signal-to-noise problems.

“Signal to noise is a major issue. Most diagnostic approaches do not even think about it or how to deal with it,” explained Lee Hood, M.D., Ph.D., co-founder and board member of Integrated Diagnostics and co-founder and president of the Institute for Systems Biology.

“Blood illustrates disease dynamics in a way that diseased tissue cannot, for you can follow disease progression and therapeutic response,” added Dr. Hood. “That is an enormous benefit in evaluating disease onset and progression. In cases where a clean phenotype separates disease from normal, blood proteins will be the disease diagnostic of the future. In the end, proteins are the biomarkers closest to the biological action.”

Three million lung nodules are identified annually, 600,000 of which fall into an intermediate category. Surgical results indicate that 50–60% of intermediate-category tumors are benign nodules, creating unnecessary system costs.

Integrated Diagnostics’ Xpresys Lung molecular-blood test provides noninvasive, objective information for assessing pulmonary nodules. The test is intended to measure the relative abundance of 13 proteins across multiple lung cancer disease pathways. These proteins, selected from almost 400 candidates, constitute a panel that can identify 60–70% of benign nodules, eliminating unnecessary surgical intervention.

The 13 proteins map into three major disease-perturbed networks for small cell lung cancer. These pathways reflect the biological disease process. After additional optimization and validation, these biomarkers may allow observation of disease progression and therapeutic response.

“I believe that in five years, blood diagnostics are going to be the key disease diagnostics,” asserted Dr. Hood. “Furthermore, these diagnostics have the ability to stratify complex diseases into subgroups to better gauge therapeutic reactions.”

Dr. Hood predicted that a new medicine is coming: “In 10 years, every patient will be surrounded by a virtual data cloud with billions of data points. We will learn how to reduce that data dimensionality for each individual to discover their internal, external, and environmental stimuli to optimize their health. In the beginning it will be crude, but with time it will become more powerful.”

According to Dr. Hood, the new medicine will be predictive, personalized, preventive, and participatory—or “P4,” to be succinct. “Molecular diagnostics will provide the tools to identify and predict disease. Systems approaches will generate preventive drugs or vaccines, plus the focus on wellness will be a powerful preventive. ‘Participatory’ is the largest challenge. All stakeholders have to accept this new transformation of medicine.”

“It is a very exciting time in diagnostics. We are right at the beginning of a revolution and it is going to be really transformational,” concluded Dr. Hood.

Integrated Genomic Analyses

Lung cancer is particularly amenable to genomic-based therapies. Work substantiating this observation has been carried out by Trever Bivona, M.D., Ph.D., an assistant professor of medicine/hematology-oncology at the University of California, San Francisco (UCSF). According to Dr. Bivona, UCSF’s Helen Diller Family Comprehensive Cancer Center has been using whole exome and transcriptome deep sequencing for nearly a year to systematically identify molecular biomarkers of response and resistance to specific targeted therapies in lung cancers.

Data are synthesized in a clinical time frame and used to direct the mechanism-based treatment of lung cancer patients, after the research-grade test results are validated in a CLIA setting.

The approach provides a comprehensive and unbiased view of the genetic architecture of each individual patient’s tumor, and can unlock the genetic secrets of tumor initiation and progression. This enables a context-specific understanding of the molecular pathogenesis along with biologically precise, individualized therapies that improve patient outcomes.

Barriers to Acceptance

Barriers slow the translation of discovery work into clinical application. Next-generation sequencing (NGS) platforms present bioinformatics challenges that the modern pathology laboratory is not equipped to deal with. And with the exception of the large teaching hospitals, there are very few community physicians that have a great degree of facility with genomics.

“If you do not understand the tests underlying the diagnostic, it is not going to be part of the evidence base that you use to make therapeutic choices. We have a huge job in front of us to train physicians so they understand the value of ’omics in the context of clinical care,” explained Elaine R. Mardis, Ph.D., professor of genetics and molecular microbiology at the Washington University School of Medicine and co-director of the university’s Genome Institute.

Reimbursement remains a confusing landscape. Some single-gene tests—for example, EGFR mutations for lung cancer—are standard-of-care and covered under insurance. Larger tests that look at multiple, hundreds, or all genes need to demonstrate clinical efficacy, and to be packaged into a palatable bite for payers. All of these parameters need to be satisfied for the successful introduction of genomics as an element for routine cancer patient care.

The Genome Institute’s cases are research studies supported by discretionary funds or grants. If something clinically relevant is found, findings are verified in a CLIA pathology lab with all the appropriate metrics to fit with the current pathology paradigm.

The institute includes RNA expression in their analyses. On average, about 40% of genes that are found to be mutated by sequencing DNA are not expressed in the RNA. RNA sequencing, although complex, has allowed, in some cases, identification of overexpressed genes for which therapeutics exist and patients have been successfully treated.

“We are still discovering what cancer is,” reflected Dr. Mardis. “We have done a ton of discovery about genes that are mutated, but the follow-on functional data to demonstrate activating mutations that drive cancer, and [to show] what the interactions are between the different pathways, is largely untouched.”

“Soon we will make or break the argument for doing genomic studies across cancer patients. The basic question is whether sequencing, at some level, provides information that is helpful for guiding patient care and results in better outcomes,” concluded Dr. Mardis.

Shutting Down Networks

A multifaceted cancer research program recently launched at Mount Sinai embraces the complexity and uniqueness of each tumor. This program seeks to understand each tumor’s entire mutational landscape, weaving DNA, RNA, CNV (copy number variant), and other information together to generate probabilistic causal networks in the hopes of identifying the perturbations that drive cancer.

“Treating single genes, or mutations, as the target is valuable but not sustainable. To increase therapeutic success and lower drug resistance, you have to start looking at networks as the target,” commented Joel Dudley, Ph.D., assistant professor of genetics and genomic sciences and director of biomedical informatics at Mount Sinai’s Icahn School of Medicine.

“Our network-modeling approach will let us build a predictive network for various tumor types. Projecting the individual patient’s information onto these networks will help identify the network where each particular tumor seems to be most active, and help determine how to target that network by downregulating or shutting it down.”

This approach may appear to be somewhat counterintuitive if one is accustomed to the current approach, which focuses on mutations and is highly targeted. All the connections and pathways that exist in biology are not fully understood, and the data-driven networks could end up looking very different what may have been imagined on the basis of common beliefs.

Various approaches are under evaluation, including second-generation, real-time, observational sequencing technology, which not only provides the sequence and long reads, but also measures kinetics. Different kinetics allows observation of epigenetic marks, and longer reads are critical for looking at structural variations, duplications, fusions, etc.

“We also look for novel epitopes. Finding an epitope that you can target on a novel gene fusion will allow vaccine design, and an immunotherapy-based therapeutic approach within an amenable timeframe,” observed Dr. Dudley. “We want to add years or more to people’s lives on average, and are taking a high-complexity, more-ambitious route because we think the payoff will be better in the long run.”

“This is a research project with translation. As we define the process, we are also operationalizing it, so when our clinical trials are completed, and we start to get evidence that this is the way to go, we can just flip the switch and start treating patients. That will be the best reward,” concluded Dr. Dudley.

The multiscale approach taken by Mt. Sinai’s Icahn Institute integrates many different dimensions of data to depict the molecular states at play in an individual. It may be possible to see how variations in these states lead to phenotypic changes related to disease or other traits of interest.

Putting Discovery to Use

Decades of cancer research have provided a wealth of information about cancer drivers, or oncogenes, and accessory cancer genes, and it is time to progress past the discovery phase. This is the motivation for the approach used by a leading research team at the Karmanos Cancer Institute to focus cytogenomic analyses, an approach that relies heavily on integrating knowledge-based evidence. One of the team’s goals is to rapidly turn around translatable insights for presently understudied cancer types such brain metastases arising from lung and breast cancers.

“There have been great attempts to integrate and curate the information from molecular cancer research. We apply this information in a comprehensive way to results of our genomic laboratory analyses to allow us to quickly interpret results from small sets of tumors, or individual primary cancer and matched metastases, to gain meaningful insights and appreciate molecular underpinnings of the tumors,” explained Aliccia Bollig-Fischer, Ph.D., assistant professor of oncology at Wayne State University and associate director of the Genomics Core at the Karmanos Cancer Institute.

“We know a lot about the biological process of metastases but little about the genomic landscape or mutation architecture of brain metastases. We collect specimens and data, analyze the tumors, and gain additional information as to the true frequencies of any individual cancer gene mutation in our patient population. We want to understand what potentially actionable oncogenes are important in the cancer subsets that we are looking at. The discovery of new oncogenes is not our emphasis.”

The very early results of this research concurs with ideas put forth already in the literature. Increased diversity of mutated oncogenes and co-oncogene expression appears to be a feature of the brain tumors arising from metastases of lung and breast cancers. Metastatic tumors arising in the brain have increased gains in amplified oncogenes compared to the primary, or original, tumors.

To improve patient treatments sooner, clinical genomic researchers at the Karmanos Cancer Institute also integrate information on existing targeted therapies and clinical trials, not necessarily limited to brain tumors, into their knowledge base to project which insights may clinically translate more rapidly.