Early Detection Research Award
Grant title (if any)
Rising Tide Foundation for Clinical Cancer Research/LUNGevity Foundation Lung Cancer Early Detection Award
This grant was co-funded by Rising Tide Foundation for Clinical Cancer Research
Abhijit Patel, MD, PhD
Yale University
New Haven
CT
Steven Skates, PhD
Harvard Medical School
Cambridge
MA
The objective of this project is to develop a blood test that can improve upon current limitations in lung cancer screening. Dr. Patel and his team have developed a method to accurately measure alterations in DNA that are cancer-specific by looking at levels of methylation of circulating tumor DNA (ctDNA) in the bloodstream. Using this method, Dr. Patel will develop a predictive model to identify patients with lung cancer based on these DNA alterations at a single time point, as well as an algorithm that can track these changes in a patient’s DNA over time. If successful, this could help detect lung cancer earlier in its development, thereby leading to better outcomes for patients.
Research Summary
Lung cancer is by far the most deadly cancer in the U.S., with total lung cancer deaths exceeding those of the next three major cancers combined. Such dismal statistics are largely attributable to the insidious nature of the disease; by the time symptoms appear, the cancer has often spread to an extent that makes cure unlikely or impossible. In contrast, patients who are diagnosed at earlier stages have much better outcomes, as their tumors can be entirely removed or eradicated prior to distant spread. Thus, annual chest CT scans for lung cancer screening have proven to be effective at reducing lung cancer deaths, and are currently recommended for patients with a heavy smoking history. However, CT-based screening programs have been practically challenging to implement, and uptake has been slow. An alternative screening approach that has been garnering much enthusiasm is based on development of a simple blood test that detects DNA fragments shed from tumor cells into the bloodstream. Several commercial and academic groups have been racing to develop blood tests for cancer screening based on this concept, and the field has made impressive progress. However, detection of early-stage lung cancers has remained particularly challenging, with sensitivities reaching only ~20-40% for Stage I disease. A key limitation for detection of small, early-stage tumors has been the extremely low abundance of DNA fragments bearing cancer-specific features (such as mutations) in the circulation. To overcome this limitation, our group has developed a technology that can accurately measure cancer-specific alterations in DNA which are more highly abundant (known as “hypermethylation”). In the current project, we propose to develop a predictive model to identify patients with lung cancer based on probabilities inferred from measurement of these DNA alterations. We will then further improve the sensitivity for detecting the earliest stages of lung cancer by developing an algorithm that tracks longitudinal changes in a patient’s DNA signal over time rather than relying on just a single time-point.
Technical Abstract
Early detection of cancer has long been one of the grand challenges of medicine. It is widely acknowledged that better methods for detection of small, asymptomatic tumors are likely to translate to substantial improvements in cancer survival rates. This is an especially important priority for lung cancer because of its high incidence, high rate of late-stage diagnosis, and high mortality. Over the past decade, liquid biopsy approaches based on detection of cancer-specific mutations or epigenetic changes in cell-free DNA (cfDNA) have made significant inroads towards this goal. However, detection of early-stage lung cancer has been particularly challenging because of the minute amounts of tumor DNA shed into blood. Methylation of cfDNA has emerged as a biomarker of choice for many early detection efforts, but existing technologies are designed to probe for cancer-specific methylation patterns either at pre-specified target sites or across broad genomic regions. The former approach prioritizes a limited subset of cancer-relevant signals, whereas the latter approach yields sparse cancer signals from extensive sequence data. Our group has developed a liquid biopsy technology that comprehensively profiles hypermethylated promoter sequences in cfDNA arising from anywhere in the genome. Using a high-stringency capture strategy based on methylation density rather than sequence, our method is able to globally profile hypermethylated promoters without pre-specifying targets. Gene silencing via promoter hypermethylation is a fundamental mechanism of carcinogenesis, and this aberrant signal can be detected at very low levels in plasma because background methylation patterns in healthy plasma are remarkably consistent. To optimize sensitivity for detection of early-stage lung cancer, we will develop a scoring scheme based on probabilistic machine learning to predict the likelihood of lung cancer by integrating hypermethylation signals across thousands of cell-free DNA fragments. Unlike most current liquid biopsy-based early detection efforts which are focused on identifying individuals with cancer based on a single time-point measurement, here we propose to develop a longitudinal early detection algorithm based on measurement of serial increases in cancer-specific epigenetic signals over time due to tumor growth and accumulating changes in the epigenome.