Liquid biopsy testing has been widely adopted to support early cancer diagnosis, therapy selection, and disease and treatment monitoring in cancer patients. The promise of minimally invasive sampling of tumor genetic landscapes to revolutionize precision medicine is driven by recent advances in next generation sequencing (NGS) methods, enabling larger panel sizes and the detection of a wider range of pan-cancer genomic and epigenomic alterations in circulating tumor DNA (ctDNA). However, genomic analysis of liquid biopsy samples presents unique challenges, both at the pre-analytical and analytical stages.
In 2023, AMP/CAP jointly published a set of recommendations for cfDNA assay validations1 based on a review of more than 1,200 publications on ctDNA assay performance in patients with lymphoma and solid tumors. Less than 50% of surveyed studies describing NGS-based ctDNA analysis studies reported LoD and only around a third reported analytical specificity. The majority, but not all, of clinical validation studies reported both. This survey highlighted lack of laboratory standards in ctDNA clinical validation literature and a need for technical ctDNA validation guidelines.
“Methods must be reliably able to detect clinically significant mutations in cfDNA at low allele frequencies, sometimes <0.5%. Detection of low-level VAFs in clinical situations requires methods that are both highly sensitive and highly specific." 1
For more details on the AMP/CAP validation recommendations and their practical applications read our blog titled Ideal Samples for Validating Next-Generation Sequencing-Based Plasma Circulating Tumor DNA Assays
LGC Clinical Diagnostics recently hosted a webinar showcasing the latest developments and applications in liquid biopsy analytical validation and harmonization using Seraseq® ctDNA reference materials. This portfolio was expanded recently by launches of first-to market methylation, lymphoma and extraction reference materials thereby empowering researchers, clinicians and assay developers to unlock the power of ctDNA analysis and easily verify assay sensitivity and performance before delivering critical patient results.
The webinar also featured an example of real-world impact of using reference materials routinely in a leading clinical lab, shared by Debbie Hughes from the Institute of Cancer Research in London. Her lab developed and validated the first pan-cancer sequencing panel optimized for liquid biopsy of pediatric tumors.
The event attracted a global audience and garnered many technical questions from the audience. While most of them were addressed live, below you can find more comprehensive answers by Krystyna Nahlik (KN) and Debbie Hughes (DH) to the most interesting ones.
In our experience of benchmarking targeted panels for ctDNA detection, commercially available reference materials are problematic because only the variants of interest are well validated and the background genotype is undefined. Is this also an issue for the Seraseq® products? Do you provide a comprehensive variant report for the GM24385 genome?
KN: This is why we use the GM 24385 background whenever possible. We do not routinely perform whole genome sequencing or whole exome sequencing for each lot of reference materials, but this is not necessary since the cell line has been characterized and sequenced numerous times, with benchmarking data available in the public domain. We do perform focused or comprehensive liquid biopsy NGS testing on each lot, to orthogonally confirm the presence and frequency of the pathogenic variants of interest and ensure that only the intended pathogenic variants are detected. We also run the same analysis of the wild type ctDNA. This data, including occasional known GM24385 germline variants which come up at around 50% VAF, is available in the technical product reports for each product lot.
I sequenced the Seraseq® ctDNA Mutation Mix v2 AF0.5% with the Universal Solid Tumor Kit for Illumina (Archer DX). Besides the listed variants, I got the list of an additional 19 Pathogenic/Likely pathogenic, Uncertain significance, or Benign/Likely benign variants. Are those variants false or true positives?
KN: The Seraseq® ctDNA v2 reference materials are designed as sensitivity controls only as they have a certain level of background noise derived from the manufacturing process. It can be observed as additional variants at VAF when interrogating VAF at 0.5% or below. If the same low frequency variants are observed over multiple runs, they are likely to be false positives deriving from the reference material and can be discounted. If different variants are observed during each run, they are more likely to be false positives as well but coming from sequencing noise. We have been reducing the false positives ever since. ctDNA complete has a slightly lower level of background noise due to improved fragmentation. The upcoming ctDNA v4 reference material, as well as existing bTMB, MRD and heme malignancy ctDNA reference materials are manufactured using a different, amplification-free, method which results in background noise being close to that of patient samples, making them suitable for applications such as MRD and as a specificity reference materials.
However, there are also true positives, which are known germline variants in the GM24385 cell line. They are usually detectable at higher allele frequencies, up to 50% and can be observed in publicly available data. Any additional variants detected in the Seraseq® products (usually with targeted ArcherDx assays) are listed on the technical documentation.
You have shown concordance data of the Seracare reference materials for liquid biopsy between assays. What about the commutability?
KN: Our reference materials can be used as a full process control to evaluate the whole workflow including extraction, to assess the effect of library prep DNA input, conversion efficiency and variant detection for all types (except copy number loss), and to establish the LoD of an assay. We do not claim full commutability, since we don’t have data directly comparing clinical sample dilution equivalent to reference material dilution across variant types and different assays working range.
None of the commercial NGS reference materials on the market can make such a claim. That would also assume NGS is being used quantitively in such a study, while in most cases it is still a qualitative method.
What is a practical way of validating and verifying LoD and measurement uncertainty (MU) of each target variant in an NGS oncology test?
KN: For this kind of uncertainty estimate, you normally need to include a large number of variants in your analysis. This can be achieved by combining several different reference materials and patient samples. Once you have a mix of both, it increases the total number of variants analyzed. However, using multiplexed reference materials offers a huge advantage of reducing the total number of samples needed.
DH: We used a similar approach and compared a large number of ctDNA sequencing samples with known variants detected from matched FFPE or fresh frozen material on our RMH200 Solid DNA NGS panel. This is a capture panel for all genes required in NHSE test directory and some additional targets considered important following literature searches. Some of them are anticipated to be added to the NHSE test directory in the near future. The assay is fully validated to ISO standards for SNV, SV, and CNV reporting, the details can be found here.
We did that across multiple tumor types as well. It is all in our paper,2 but I think the first run pool has to be a control because if you have a problem and you are not detecting the variants in ctDNA that you have identified in FFPE, you need to know that it's not an issue with your assay performance. You need to confirm that it is fit for purpose, that it is covering all the relevant regions and capable of picking a mutation up that is actually present in ctDNA. Obviously, that brings you back to the key problem with ctDNA analysis: is there actually any ctDNA in there, or is it just all cell free DNA from normal cells? The tumor heterogeneity was an issue as well, especially in in case of neuroblastomas which often have a lot of metastatic sites.
To sum up, to verify LoD and measure the uncertainty you need large sample pools and real-life samples but the controls are the best place to start.
To watch the webinar, click here or on the button below to gain access.
References: