NEW YORK (GenomeWeb) – It may be difficult or impossible to distinguish authentic intra-tumor heterogeneity from sequencing noise and technical artifacts with whole-exome tumor sequencing done at standard depths of coverage, according to a new study in Cell Reports.
Researchers from Yale University, Memorial Sloan Kettering Cancer Center, and elsewhere did exome sequencing on biopsies from three anatomically distinct regions per tumor for six individuals with primary breast cancer. Their analyses of the tumors, along with matched normal blood samples from the same individuals, indicated that some 69 percent of predicted somatic mutations identified by these analyses alone were false positive differences reflecting technical noise or misidentified germline variants.
The analyses hinted that the accurate detection of intra-tumor somatic mutation differences could be bumped up by increasing sequence coverage and/or leaving low-mappability regions of the genome out when analyzing the exome sequence data, the team noted, though more aggressive filtering for minimum sequencing depth did not appear to add much to the somatic mutation detection sensitivity.
Co-corresponding authors Christos Hatzis and Jorge Reis-Filho — researchers affiliated with Yale and MSKCC, respectively — and their colleagues noted that intra-tumor genetic heterogeneity is an "important consideration in the clinical setting" given the impact it can have on a patient's potential prognosis and predicted treatment response.
"Because [of] the cost of [whole-exome sequencing] and complex informed consent requirements, the inclusion of matching normal samples still represents a limitation in the sequencing of tumor samples in some large clinical trials," the authors noted. "Whether sub-clonal mutations can be robustly identified in the absence of matching normal samples and whether pooled normal samples from unrelated individuals would serve as a reasonable control should be explored as alternative options for the assessment of [intra-tumor genetic heterogeneity] in the clinical setting."
To explore the reliability of intra-tumor somatic variant detection with or without accompanying matched normal sequence data, the researchers considered three samples apiece from four estrogen receptor-positive breast cancer cases and two cases involving triple-negative tumors that lacked estrogen receptor, progesterone receptor, and HER2 expression. They used Illumina HiSeq 2000 instruments to sequence protein-coding parts of the genome that had been captured from the samples with NimbleGen SeqCap human exome kits to a mean depth of 184-fold in several technical replicates each.
After identifying potential somatic variants and small insertions and deletions with three independent analytical pipelines, the group followed up on apparent mutation differences within each tumor with the help of high-depth amplicon sequencing at 605-fold median depth. With the validation step, the availability of technical replicates, and sequence data from matched normal samples sequenced to 90-fold coverage, on average, the researchers evaluated the possibility of finding genuine intra-tumor genetic variants with these approaches.
From these and other analyses, the authors concluded that "[whole-exome sequencing] performed at typical sequencing depth may be inadequate for detecting [intra-tumor genetic heterogeneity], particularly when tumor cell content is less than 50 percent, as only 62 percent … of the somatic mutations were detected consistently in the technical replicate pairs by [whole-exome sequencing], with the remaining mutations falsely appearing as discordant."