Skip to main content
Premium Trial:

Request an Annual Quote

Lung Cancer Liquid Biopsy Approach Marries Sequencing, Target Enrichment, Computational Methods


NEW YORK (GenomeWeb) – Researchers continue to garner new funding to prove that DNA signals in the blood can catch cancer cases before other symptoms develop, including now a Yale University-led group advancing its efforts in non-small cell lung cancer.

Headed by Yale assistant professor Abhijit Patel and Rice University bioengineer David Zhang, the team is one of six that have been granted five-year funding under a National Institutes of Health program focused on establishing academic-industrial partnerships toward development of tests for cancer early detection and monitoring.

Patel and colleagues received a little more than $500,000 last month to begin their first year of work, and similar amounts are supporting five other projects, which include two efforts at Massachusetts General Hospital, one of which is developing exosome-based methods to assess early-stage brain cancers and the other assessing exosomes for use in ovarian cancer screening.

A third team led by University of Miami researchers is developing a circulating tumor cell imaging technology for predicting recurrence in early-stage breast cancer patients. A fourth, at Johns Hopkins, is expanding on earlier findings in Pap smears to develop liquid biopsy tools that can detect cancer in cerebrospinal fluid, saliva, and stool samples. And a fifth group, at the University of California, Los Angeles, is working with an electrochemical technology called EFIRM, hoping to combine genetic and microRNA targets to detect early-stage lung cancers.

For their part, Patel and Zhang are hoping to prove that DNA sequencing error-correction and noise-reduction technologies they have each been working on can combine to help detect the earliest signs of lung cancer in circulating cell-free DNA. By the end of the five years of funding, they plan to have data that demonstrates how their approach has performed in samples collected prospectively from a group of high-risk individuals being screened for lung cancer.

In many ways, the goals and the approach are similar to what has been disclosed by commercial firms in the space like Grail and Freenome.

Patel said that his group is employing artificial intelligence, or machine learning technologies — in this case working with Microsoft — to help narrow and optimize a DNA-based signal that best discriminates early cancers from normal control samples in a way that would work to screen an asymptomatic population.

Patel and his group at Yale have previously reported on two methods — one for paired-end sequencing, and the other using molecular barcoding — to help reduce DNA sequencing errors, making it easier to catch the trace fragments of DNA necessary for early cancer detection.

In earlier work, the Yale team's library preparation and error suppression efforts have been amplicon-based, but because early cancer detection presumably requires broader mutation coverage, Patel said that they have transitioned over the last two years to ligation and hybrid-capture.

With Zhang, they are now adding the toehold probe technology he created at Rice, which helps eliminate wild-type DNA in favor of mutated targets before amplification and sequencing.

The overall goal, Patel said, is to apply these technologies in an initial cohort of blood samples from patients with early-stage tumors who had them surgically removed and a set of normal controls — using the machine learning capabilities at Microsoft Research Lab in Cambridge, UK, to help define a signal with the best possible specificity and sensitivity for distinguishing the two. He added that the approach will involve measurement of both overall genomic methylation and a specific panel of gene mutations.

Comparing various efforts in the liquid biopsy early detection space is difficult when groups, and especially companies, have released little information on their approaches. However, Patel said that part of what he and his colleagues believe to be unique to their approach is the efficiency and cost savings afforede by their technologies.

Patel and colleagues have previously reported that they were able to create a circulating tumor DNA assay with an error rate of less than one in 100,000 bases, with a sequencing cost of $20 per ml of input plasma.

Zhang's toehold probe method will also help streamline the team's assay, Patel added. "We want to make this as affordable as possible so being able to decrease NGS costs by depleting wild type sequences from the library" is crucial, he said.

Importantly, the assay approach involves molecular tags introduced up front, and even though allele fraction information disappears when the wild-type DNA is depleted, the group can still count molecules and retain quantitative information, he added.

Over the five years of the grant, the group intends to collect appropriate clinical data to validate the assay. Investigators will also do a preliminary prospective analysis, using the test in a cohort of non-cancer patients who are being screened for lung cancer using radiologic imaging. "We'll probably be able to collect a few hundred samples, which means there will only be a handful of positives … [so it will be small scale] but it will be a good proof of principle," Patel said.

Harvard Medical School biostatistician Steve Skates, a third member of the project, and a principal investigator in the NCI's Early Detection Research Network, will be lending his expertise in designing and implementing the clinical validation arm of the effort.

According to Patel, all the teams funded under the liquid biopsy early detection RFA are working together as a consortium and will be meeting regularly, so there will be opportunities for the groups to share samples, "and maybe also technologies," he added.