NEW YORK – Precision oncology firm Burning Rock Biotech this week presented data at two scientific conferences supporting two products it expects to be major growth drivers in coming years.
At the annual meeting of the Association for Molecular Pathology, held virtually, researchers from the company presented a study describing the validation of its automated next-generation sequencing library prep solution Magnis BR, which was launched last year.
Separately, Burning Rock investigators also shared new data on the implementation of the firm's ELSA-seq technology for liquid biopsy cancer screening at the European Society for Medical Oncology Asia Virtual Congress 2020. The new results both extended the test to three new tumor types and showed a significant specificity improvement compared to earlier versions.
Burning Rock is currently in the process of evaluating ELSA-seq for early cancer detection in what is intended to become a 14,000-patient trial called the Pan-Cancer Early Detection Project (PREDICT), which would support a commercial launch of an assay targeting nine cancer types: lung, colorectal, liver, ovarian, pancreatic, esophageal, gastric, head and neck, and cholangiocarcinomas.
Like Grail, Freenome, and Guardant Health in the US, and Genetron, Singlera, and others in China, Burning Rock's cancer early detection approach relies on the detection of tumor-associated epigenomic signals in circulating cell-free DNA.
At the ESMO Asia meeting this week, investigators reported on a study using lung, colorectal, liver, ovarian, pancreatic, and esophageal cancer samples, mostly at early stages, in which ELSA-seq achieved about 98 percent specificity and nearly 81 percent sensitivity.
During a call discussing the firm's quarterly financial results, Burning Rock's Chief Technology Officer Joe Zhihong Zhang described the newly presented data, which he said included three stages: a biomarker discovery study, validation of the resulting epigenetic target panel, and assay validation.
To develop the six-cancer detection assay, the company analyzed cancer tissue samples and also surveyed public databases to identify "key methylation changes associated with the six prespecified cancer types," Zhang said.
The result was approximately 160,000 CPG sites, which researchers then used to construct a customized capture-based panel. They also incorporated a machine learning classifier to enable localization of cancer-derived signals, or tissue-of-origin prediction.
To validate the assay, the company used a set of ctDNA samples from clinically diagnosed cancer patients and non-cancer controls, which investigators randomly divided into independent training and validation cohorts.
Zhang stressed that both the case and the control group were similar with respect to age, gender and smoking status. "As we know methylation varies … and is an age-related biomarker. We've been really careful about … making sure [these confounding factors] are controlled," he said. About 80 percent of the overall cohort was also made up of cancers diagnosed at stage I-III.
According to Zhang, the performance was consistent across the training and validation sets, which lends confidence that the results may hold up in clinical validation in the ongoing PREDICT trial. In the training cohort, specificity was 99.5 percent and sensitivity was 79.9 percent. In the validation set, specificity dropped slightly to 98.3 percent, while sensitivity rose to 80.6 percent.
Detection was highest for colorectal and liver cancers, and lowest for lung tumors, and was lower for the earliest stage cancers than the more advanced cases.
Finally, the tissue-of-origin classifier yielded a result for almost 99 percent of the samples, with about 82 percent accuracy using a single-organ call, or 87 percent when allowed to predict the two most likely origins.
Shannon Shaokun Chuai, Burning Rock's chief operating officer, added that the test's performance in the current study, although limited in some tumor types and at earlier stages, bodes well for adoption, especially considering the fact that China does not have the same type of existing cancer screening systems in place as seen in the US.
"We actually think this is already a very good performance for the early detection market especially given that it's a multi-cancer assay," she said during the call.
"Keep in mind that with multi-cancer we have to balance between sensitivities … so this is what our research team has so far figured out as to the best balance. But of course, there are also ... [improvements] possible in the future, especially since this is still a relatively small cohort," she added.
Zhang added that many other companies targeting the Chinese cancer screening market had focused exclusively on liver cancer, and the company's performance for those tumors in the study was competitive, even against these single-tumor tests.
Chuai declined to provide any updates on the larger PREDICT trial.
Cancer early detection is only one business area Burning Rock is hoping to expand in coming years. Another goal has been to grow its business outside of the in-house sequencing services that currently provide most of its revenue, by providing disseminated sequencing kits to hospital labs.
To support this arm of its business, the company developed what it calls Magnis BR, an automated NGS library preparation system, which it launched last November.
According to Burning Rock, hybrid capture-based NGS library preparation is a hurdle in disseminated NGS test adoption, because the process is time consuming and requires specialized skills that many community hospital labs may struggle with. The Magnis BR platform is designed to create libraries automatically without the need for hands-on attention and in as little as 9 hours compared to the 24 to 36 hours required by manual approaches.
In a presentation this week at the Association for Molecular Pathology meeting, researchers described a performance validation of the system using its 520-gene panel OncoScreen plus tissue test and 168-gene liquid biopsy panel, which they said demonstrated high sensitivity, precision, specificity, and accuracy, with improved detection of microsatellite instability status.
According to the authors, Magnis BR improved the overall capture-on-target ratio by an average of 8 percent compared with manual library preparation when applied to a series of contrived samples.
The sensitivities for detecting SNVs, indels, and rearrangements all reached 100 percent at an allele fraction of 2 percent, as did detection of copy number variation and microsatellite instability at respective relevant frequencies. In a normal control sample, the method yielded no false positives.
The group also analyzed clinical FFPE and liquid biopsy clinical samples. In 16 tissue samples, the detection rate was 98.7 percent across 717 analyzed variants, and 100 percent for detection of MSI-high signals. In 34 clinical cfDNA samples, the concordance rate across 120 variants was 97.5 percent.
Magnis BR also detected a higher proportion of MSI sites than manual library prep in the MSI-high samples.
Burning Rock said that the demand for Magnis BR has been strong and that the system has now been placed in more than 10 hospitals.
According to Zhang, the platform is also now compatible with several of its NGS products, including its 168-gene liquid biopsy assay and its 520-gene tissue-based NGS panel.