NEW YORK (GenomeWeb) – Researchers at the University of Michigan and Thermo Fisher Scientific have developed and validated a comprehensive next-generation sequencing-based cancer panel on Thermo Fisher's Ion PGM.
The Oncomine Comprehensive Panel, which evaluates both DNA and RNA, is being used in the National Cancer Institute's Match trial, a basket trial that seeks to stratify patients into clinical trials based on their genomic profiles.
Reporting in the journal Neoplasia, the researchers reported that the panel was a "rapid, highly scalable approach" with "highly concordant results with typical specimens sent for molecular diagnostic testing and molecular standards."
To design the panel, the team mined the Oncomine database, which includes mutation, copy number, and gene fusion data from over 700,000 cell lines, xenografts, and clinical cancer samples, to identify relevant alterations.
The team first developed multiple DNA and RNA panels for a final design that totaled 143 cancer genes including 73 oncogenes, 49 copy number altered genes, 26 tumor suppressor genes, and 22 fusion driver genes. The targeted DNA panel included 2,530 amplicons covering 260,717 base pairs in 130 different genes. The targeted RNA-seq panel included 154 primer pairs targeting 148 known gene fusion isoforms, as well as 5' and 3' expression assays for RET, ROS1, and ALK to enable novel fusion discovery.
In addition, a "critical component" of the panel is the analysis pipeline that links to a knowledgebase of potential treatment options, the authors wrote.
The authors also noted that the DNA and RNA components can be combined for template preparation and sequencing on the PGM using the 318 chip in a four-hour run. Another advantage of the assay, they noted, is its low input requirements — 20 ng of DNA and 15 ng of RNA.
To validate the panel's performance, the group assessed the DNA component using a cell line engineered to contain 398 variants on the panel at 0.2 expected variant allele frequencies. The panel identified 364 out of 365, or 99.7 percent of SNVs, and 25 out of 33, or 75.8 percent, of indels. Of the eight indels that were missed, three were over 10 bases in length and five were single nucleotide insertions or deletions that occurred within a homopolymer run. "Accurate indel identification in homopolymer runs is a known challenge with current Ion Torrent sequencing technology," the authors wrote.
Next, they looked at DNA from a cell line mixture with mutations at various precise variant frequencies. In total, the cell line contained 16 known mutations that were targeted by the Oncomine panel. Applying an automatic variant caller, they were able to detect 15 out of the 16 variants. The one missed variant was a secondary frameshift deletion at the start of a homopolymer run that was present at a frequency of 7.5 percent.
The group then validated the panel on formalin-fixed paraffin-embedded tissue samples from three cohorts: 105 samples sent for routine molecular diagnostics, 104 samples with retrospective lung cancer, and 118 prostate cancer samples. About one-third of the samples were at least three years old and six samples could not be analyzed due to low-quality libraries.
On average across all three cohorts, the DNA panel obtained 5,142,690 mapped reads, 97 percent of which were on-target; 1,941x coverage across targeted bases; 93.6 percent of targeted bases covered by at least 20 reads; and 202 called variants per informative sample. For the RNA panel, the group obtained an average of 306,872 mapped reads, including 210,712 reads that mapped to the five housekeeping genes.
Finally, the group validated the panel in a clinical cohort, assessing 105 FFPE samples sent for clinical molecular diagnostic testing for EGFR, BRAF, KRAS, and ALK alterations. The samples included 29 colorectal adenocarcinomas, 23 lung adenocarcinomas, 48 melanomas, and four other cancers. They identified an average of 1.7, 0.8, and 1.7 relevant SNVs, indels, and CNAs, respectively, per sample. The most frequently mutated genes were TP53, BRAF, and APC, which were mutated in 33 percent, 31 percent, and 24 percent of the samples, respectively. They also identified four gene fusions, including multiple isoforms of the ERC1:BRAF fusion.
In total, the DNA component of the panel had 100 percent sensitivity and specificity for detecting the clinically identified EGFR, BRAF, and KRAS mutations. For the RNA component, the panel was 100 percent concordant for T2:ERG fusions, and identified five out of seven, or 71 percent, known ALK fusions. One of the missed fusions was detected but below the threshold for automated calling. In addition, the panel identified two additional relevant fusions.
As clinical sequencing becomes more prevalent, a "key advantage of the OCP is the potential for integration into multiple independent institutions (rather than a single centralized testing center), enabling valuable direct involvement from molecular biologists, pathologists, and oncologist," the authors wrote. In addition, the assay may have "utility in future oncology precision medicine approaches, such as the NCI Match Trial, where multiple sites will sequence 3,000 cancer samples."