NEW YORK (GenomeWeb) – Fluxion Biosciences is using a $300,000 grant from the National Cancer Institute to further validate its liquid biopsy-based, mutation variant caller technology for sequencing services.
The firm has partnered with the Cleveland Clinic to validate its "elimination of recurrent artifacts and stochastic errors" sequencing (Erase-Seq) technology — which separates out the sources of error-systemic biases and stochastic errors using technical replicates and background error modeling— in early-stage hepatocellular cancer samples.
Fluxion was founded in 2006 in San Francisco to commercialize microfluidic research developed by Cofounder and Chief Technology Officer Cristian Ionescu-Zanetti at the University of California, Berkeley. Fluxion CEO Jeff Jensen noted that the firm has since moved to Alameda, California and is licensing approaches for circulating tumor cell (CTC) enrichment and analysis from the university.
Jenson explained that Fluxion previously looked at other mutation caller techniques to integrate with its IsoFlux CTC isolation kits and Spotlight liquid biopsy panels, but the firm found their performance "unacceptable." He argued that the lack of high-quality callers led the group to develop and eventually release Erase-Seq in 2018, which is more of a statistical approach to identifying the true positives in a sample.
According to Jensen, Erase-Seq borrows techniques researchers would apply in a signal-to-noise limited environment during physics experiments.
A researcher initially performs next-generation sequencing on a prepared sample and generates FASTQ data. Erase-Seq then identifies systemic biases and sequencing read errors, which are two types of errors that often show up in the resulting data after NGS.
Using custom error models to mitigate the two issues, the system applies statistics to look across the replicate population and establish a consensus read and determine the read's validity. From the statistical analysis, the team can produce a confidence score for every read at single position in the panel.
"Some spots [in the genome] might have inherently louder noise than other spots, but we're able to account for that," Jensen said. "Rather than using a fixed cutoff, we use a probabilistic cutoff whether the confidence is likely to be real or an error."
In a study published in PLOS One in April 2018, Ionescu-Zanetti and his colleagues compared the abilities of ERASE-Seq and those of current molecular barcoding methods. Using spiked human DNA mixed with clinically realistic DNA input quantities, the researchers sought to detect single nucleotide variants and indels between 0.05 percent and 1 percent allele frequency. They found that the tool could detect variants with greater than 90 percent sensitivity and a false positive rate below 0.1 calls per 10,000 possible variants.
While Erase-Seq can be used in tandem with the IsoFlux CTC assay and Spotlight NGS panel, Jensen said that more research groups have frequently used Erase-Seq for ctDNA analysis. He explained that despite demonstrated interest in CTCs in the clinical research side, many researchers are looking at ctDNA because the analyte doesn't require sample enrichment.
Jensen said that researchers have used Erase-Seq to identify ctDNA and CTCs in several cancer types, including lung, colorectal, melanoma, and prostate cancers. He noted that his group can develop an algorithm that will customize the Erase-Seq approach and chemistry for any type of cancer.
According to Jensen, Flexion can validate a lab's panel using the Erase-Seq variant caller approach. The validation involves running a set of reference healthy samples through the firm's algorithm to develop a background model. Afterwards, the researchers run a few analytical samples to adjust the sequencing parameters, such as the depth of coverage.
In terms of time to result, Jensen said that his team currently has a file loader capability, where it uploads the files to servers it has on Amazon Web Services (AWS). The actual number crunching of data occurs on AWS, which he said takes "overnight."
In addition, the team is also building a second approach that allows researchers to upload their data onto a local server. Jensen said that this method will still require overnight to produce a tailored report.
"We know that some labs ultimately don't want the data leaving their premises, and we therefore are developing an automated approach that lets labs perform the test on their site," Jensen explained.
As part of its NCI Phase I Small Business Innovation Research grant, which began in September, Jensen and his team received $300,000 to further develop Erase-Seq for potential downstream clinical analysis. According to the grant, the team will use the funding to achieve three major aims: to improve Erase-Seq's bioinformatic algorithms in order to increase its clinical sensitivity and specificity, to automate the entire Erase-Seq data analysis pipeline, and to compare the sensitivity and accuracy of the platform to other sequencing approaches using cancer patient blood samples.
As part of the ongoing study, Cleveland Clinic assistant professor of radiation oncology Omar Mian and his colleagues are applying Erase-Seq to track tumor growth in a group of 30 hepatocellular cancer patients. The team is attempting to see whether the circulating tumor burden of disease and the amount of ctDNA changes before, during, or after ablative therapies (such as stereotactic radiation), surgical management like mastectomy, or tumor removal.
"Our plan is to set up studies doing blood draws, then compare analyses run on our platform and other platforms, and then use samples for a third orthogonal test," Jensen said. "Right now, we're planning to use lung cancer samples, and then use [Bio-Rad] digital droplet PCR for our orthogonal test, which we consider the best standard we have right now."
If successful, the group believes the platform will help variant detection in cfDNA samples and offer improved sensitivity for a variety of workflows and panels without issues linked to unique molecular identifier (UMI) tags.
While other variant callers for liquid biopsies use UMIs to reduce noise, Jensen argued that the barcode reactions themselves are noisy and don't perform well at low molecular allele frequencies (MAF) thresholds.
"One of the challenges with [precision medicine] is the ability to call at low mutant allele frequencies," Jensen noted. "The technology we've seen, which are barcoded approaches, start to have specificity issues below 0.5 percent."
At the same time, Jensen said that researchers typically need to call several mutational variants that have an allele frequency of 0.5 percent or lower. The company's goal is to therefore push the limit of detection further down to as low as 0.01 percent in order to call the variants accurately in early-stage patients and potentially start intervening before metastasis occurs.
Jensen claims that using inter-sample signal processing for Erase-Seq (statistically analyzing replicates to a highly characterized background model) instead of intra-sample quality testing improves sensitivity and specificity in MAF range below 0.5 percent.
Jensen explained that Fluxion had been in discussions with the Cleveland Clinic before it began collaborating with Mian's team on hepatocellular carcinoma. Mian believes that his team could use the method for early detection and precision diagnosis of the disease.
Jensen said that his team hopes to finish the current Phase I study within the next six to months, which would potentially lead to a Phase II clinical study.
According to Jensen, Fluxion received a patent for the mutation caller technology in January. Fluxion is currently offering the Erase-Seq service for research use only, but he said it is currently in discussions with the US Food and Drug Administration for in vitro diagnostic use. Jensen believes that Erase-Seq will ultimately be decentralized and that clinical groups will eventually integrate the platforms as part of their own laboratory workflow.
In addition, Jensen highlighted that researchers can adapt targeted panels to Erase-Seq with software validation. So far, currently validated panels include Fluxion's own Spotlight 59 assay, Illumina's Trusight Tumor 15 assay, and Swift Bioscience's 56G.
Mian noted that using a reliable detection approach like Erase-Seq will be critical to eventually individualize patient treatment.
"Our hope is that we're improving sensitivity by timing the collection of liquid biopsy specimens to coincide with the point that the circulating volume tumor DNA is at its highest," Mian said. "We need [an effective way] to detect the emergence of resistance, monitor disease burden and response to therapy, as well as minimal residual disease after therapeutic intervention."