New Tools for Achieving MRD Detection, Accuracy

February 23, 2022

By Allison Proffitt

February 23, 2022 | On the first morning of the Molecular & Precision Med Tri-Con event, Viktor Adalsteinsson, associate director of the Gerstner Center for Cancer Diagnostics at the Broad Institute, explored new tools and techniques for detecting minimal residual disease.

The significance of detecting MRD will vary by cancer type and context, but Adalsteinsson emphasized that residual cancer detection can allow physicians to intensify therapy before relapse when there are presumably fewer drug-resistant tumor cells in the body, determine early if treatment is working, and deescalate therapy when MRD is cleared.

“As efforts for cancer screening ramp up, we think it’s going to be more and important than ever to detect MRD and use it to optimize the right amount of treatment for each patient,” Adal said.

While the idea of MRD originated in hematological cancers, Adal highlighted recent advances to expand it to solid tumor cancers including breast, colorectal, lung, and melanoma. Clinical trials are underway now, he said, to test MRD diagnostics and determine if acting on those findings significantly changes patient outcomes and cancer care.

But there are still patients for which MRD is difficult to detect. “We therefore believe there is an outstanding need for higher sensitivity, and that improving the sensitivity of liquid biopsy that detect MRD should be a major focus of ours.”

Adal believes it is conceptually feasible to detect one part per million of circulating tumor DNA in blood, but at that concentration, he argues that we need a better catalog of tumor mutations with which to identify the circulating tumor DNA. In fact, given the genetic diversity of most patients’ tumors, he advocates for individualized assays. By identifying 488 single nucleotide variants unique to each patient’s tumor, he predicts that individualized assays would be accurate down to 1ppm for most patients.

The Most Mutations

Theorizing that it will never be possible to detect MRD at levels lower than the sequencing error rate, Adal and his team conducted duplex sequencing to confirm any mutations by sequencing both strands of DNA. This does represent a significant cost increase over standard NGS sequencing, he noted.

The team tested the individual assays on 142 early-stage breast cancer patients using whole exome sequencing (WES) of each patient’s tumor. They found a median of 57 mutations per patient to track in plasma. “We found MRD detection post-op was highly predictive for future distant occurrence, but with the highest predictive power being observed at one year after surgery after completion of adjuvant chemotherapy,” he said. But the MRD assay did not predict all recurrence.

The group is now pursuing whole genome sequencing of the tumor, developing a test for 1,000 tumor mutations, and applying that list of mutations to a test of cell-free DNA. “We think that this is not more feasible than ever, given how cheap DNA sequencing has become,” he explained. He predicts that tumor genome sequencing will one day replace targeted gene panels and sees additional uses for the sequencing results beyond MRD including treatment selection.


The team is now focusing on validating their sequencing technique starting with sample preparation to get the most accurate test possible. They validate tests against both tumor and germline DNA and have achieved an error rate of 1 in 10 million. They have explored how the volume of blood impacts test accuracy. While high volumes of blood may be infeasible routinely, Adal suggests its use right after surgery, before stopping adjuvant therapy, or when MRD detection is on the cusp of detection.

The team has also developed a mutation enrichment method called MAESTRO: minor allele sequencing through recognition oligos (in press, Nature Biomedical Engineering). The approach relies on short allele-specific oligos under optimized thermodynamics to enable enrichment of rare, mutated DNA molecules and their detection with minimal sequencing.

The method has been used in a variety of contexts, Adal says, and the team has found that genome-wide MAESTRO assays yielded comparable results to the enhanced duplex sequencing results.

Cutting Cost

While MAESTRO does enrich mutations for testing, Adal argues that there is still a weakness in sequencing itself: platforms sequence only one strand of DNA at a time, and the duplex sequencing is expensive.

The solution, he posits, is CODEC sequencing, sequencing both strands of DNA together as together as a single strand (bioRxiv DOI: 10.1101/2021.06.11.448110). It provides the fidelity of duplex sequencing with substantially fewer reads, saving cost, he says. In addition, re-reading the CODEC strands does not increase accuracy. CODEC is compatible with most NGS workflows, Adal says, and could be used any many useful ways in the future, for example for direct sequencing of cfDNA duplexes.

Finally, the team developed Duplex-Repair, a sample prep technique that limits the propagation of DNA-based damage to both strands. Using both Duplex-Repair and CODEC can improve sequencing accuracy to 4-5 fold.