August 2015, Vol 4, No 4
Predictive Potential of Molecular Biomarkers Promises to Enhance Precision Medicine in Oncology
Recent advances in predictive and prognostic molecular biomarkers promise to enhance precision medicine in oncology. A rundown of the progress being made in bioinformatics, genetics, and proteomics for this purpose was delivered by several speakers at PMO Live 2015.
In discussing bioinformatics, Michael Kattan, PhD, argued for a deemphasis of hazard ratios in evaluating novel markers (ie, less reliance on correlation with established markers). The hazard ratio of novel markers can be affected by coding of new markers and established markers, which established markers are included in a multivariable analysis, and the modeling of established markers, said Kattan, Professor of Medicine, Epidemiology and Biostatistics, Cleveland Clinic Lerner College of Medicine of Case Western Reserve University, OH.
Perfect prediction from a single marker is unlikely, he said. Instead, a useful marker is one that improves the accuracy of a model that lacks the marker. Therefore, the best way for new markers to be incorporated into risk prediction is to evaluate their incremental predictive accuracy to established models, and not in isolation of a model. “Does marker X help us predict outcome beyond what we can achieve using existing markers…have you really exploited the existing markers,” he asked.
“Prediction is accomplished by a model of markers, not just the markers themselves,” Kattan said. A proposed solution is to calculate the improvement in the concordance index associated with a new marker, check calibration of the model with a new marker, show a scatterplot of model predictions with versus without the new marker, and perform a decision curve analysis.
The ultimate goal for evaluation of new markers is to assess their impact on clinical decision making, rather than potentially look at small differences in prognosis, he said.
Katherine Hoadley, PhD, followed with an overview of cancer genomics, noting that one of the keys in assessment of risk is to evaluate genes of importance rather than gene mutations with functional consequence, whether both exist together, separately, or are intertwined.
Genetic testing for high-penetrance germline genes is warranted in patients with a family history of cancer, whereas testing for “low effect size” germline susceptibility is not, she asserted. Hoadley used the example of breast cancer, for which BRCA1/2 germline variants increase the risk by 15- to 20-fold, whereas other genetic variants identified confer much less risk.
Somatic mutation whole genome sequencing and gene panels for clinical samples are in the early stages and still need more evaluation for genes/mutations and mutations across tumor types.
Genomic signatures versus single genes have been developed that can predict risk or response to therapy, said Hoadley, Department of Genetics, Lineberger Comprehensive Cancer Center, University of North Carolina at Chapel Hill. Though few are currently FDA approved, more are under development.
Inhibiting one part of a pathway may not be effective if other feedback mechanisms or mechanisms of resistance exist. Therefore, an improved understanding of genetic mutations and other pathway alterations (ie, copy number changes, DNA methylation) in cancer that represent potentially actionable therapeutic targets is needed.
Towia Libermann, PhD, introduced proteomics in the search for prognostic and predictive markers. Limiting the development of molecular diagnostics is the sheer number of proteins and posttranslational modified normal and cancer proteins that can be evaluated. There are hundreds of thousands of protein isoforms and millions of posttranslational modified normal and cancer proteins to complicate the discovery process.
Cancer is a systems biology and network disease with dynamic proteome alterations, said Libermann, Director, BIDMC Genomics, Proteomics, Bioinformatics and Systems Biology Center, Beth Israel Deaconess Medical Center, Boston, MA. Opportunities for protein biomarker discovery lie in tumor heterogeneity and evolution, genetic and epigenetic alterations, complex and dynamic protein interactions, and multiple cellular processes, among others.
A technique called SOMAscan, a highly multiplexed, sensitive quantitative proteomic tool, is an emerging platform for protein biomarker discovery. “It measures simultaneously 1129 proteins per sample in only 65 µL of serum,” he said. At the heart of the platform are protein-capture SOMAmer (Slow Off-rate Modified Apatmer) protein-binding reagents, which consist of a unique short DNA sequence that incorporates several bases that have been modified chemically.
Another emerging platform for protein discovery is the low multiplexed (up to 10 proteins) Quanterix Simoa HD-1 Analyzer, which is a digital immunoassay platform. It is 1000 times more sensitive than ELISA, he said. With its high sensitivity, it has high potential for detection of new cancer biomarkers in the blood or urine at ultra-low levels. The caveat is that it’s not a global proteomics strategy and therefore has the need for a preconceived notion of candidate proteins.
Mark Sausen, PhD, spoke about methodologies for biomarker testing in oncology, including a description of the potential of circulating tumor DNA (ctDNA) for both early detection and monitoring of cancer and detection of resistance mutations. The ctDNA is detectable across tumor types with rearrangements and sequence mutations, he said.
Many different types of clinically actionable somatic genetic mutations have been discovered, including sequence mutations, copy number alterations, and translocations, an increasing number of which are targetable biomarkers, said Sausen, Vice President of Research and Development, Personal Genome Diagnostics, Baltimore, MD. In non–small-cell lung cancer (NSCLC), targeted therapies approved and recommended in National Comprehensive Cancer Network guidelines are erlotinib, afatinib, or crizotinib, depending on the results of molecular testing.
Evaluation of matched tissue and plasma concordance for detection of somatic alterations reveals about 90% concordance between alterations identified in the tissue compared with those identified in plasma using noninvasive liquid biopsy approaches.
He provided an example of a patient with NSCLC treated with epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor (TKI) therapy in whom ctDNA was analyzed from a 6-mL sample of plasma. The plasma was sequenced to 41,636-fold coverage, and a sensitizing EGFR exon 19 deletion was discovered. The ctDNA was able to detect the presence of acquired resistance EGFR p.T790M mutation, a secondary mutation that occurs in about 60% of patients with acquired resistance to EGFR TKIs. The liquid biopsy thus indicated tumor recurrence as well as therapeutic resistance to targeted therapy.
Three distinct, mutually exclusive subtypes of leukemia stem cells (LSCs) have been identified, and these are correlated with specific cytogenetic/molecular risk factors and are also correlated with response and outcomes. LSCs comprise less than 1% of cells at diagnosis of acute myeloid leukemia (AML), and these stem cells are thought [ Read More ]
Targeting genetic markers in premalignancy is an emerging concept. In speaking at PMO Live 2015, Scott M. Lippman, MD, said that genetic drivers can identify premalignant conditions and even certain benign conditions, and genetic drivers can aid in identifying higher-risk populations and populations most likely to respond to targeted agents. [ Read More ]