Tumor-Only Sequencing With a Suspicious Variant Allele Fraction
A 58-year-old woman with metastatic lung adenocarcinoma undergoes targeted DNA sequencing of a 60-gene panel on formalin-fixed paraffin-embedded tumor. Estimated tumor content by pathology review is 45%. The report calls a pathogenic EGFR exon 19 deletion at a variant allele fraction (VAF) of 38%, and also flags a truncating variant in a DNA mismatch repair gene (PMS2) at VAF 47% with no matched normal. The oncologist asks whether the PMS2 finding represents a germline Lynch syndrome diagnosis. The laboratory's validated limit of detection is 5% VAF at 500x coverage; this amplicon had 180x. The patient has no family history reported, and no prior germline testing exists.
1. Interpret the EGFR VAF in the context of tumor purity and explain what it implies about clonality and copy number.
Read the worked answer
With ~45% tumor content and ~38% VAF, the observed VAF is close to what a heterozygous clonal variant would show if tumor fraction were near 76% (VAF ≈ tumor fraction × 0.5 for a heterozygous event). Because 38% exceeds 0.5 × 45% = 22.5%, the simple heterozygous model is insufficient. Two explanations dominate: (1) the tumor content estimate is low (common with desmoplasia, necrosis, or inflammatory infiltrate), or (2) the variant lies on an amplified/LOH region or the locus is subject to copy-number gain. A VAF substantially above the purity-adjusted heterozygous expectation suggests either copy-neutral LOH with the variant retained, or amplification of the mutant allele. Clinically, a high-VAF EGFR exon 19 deletion is consistent with a clonal, truncal driver and supports EGFR-TKI eligibility. The teaching point: VAF must always be interpreted against tumor cellularity and local copy number, not read as a fixed 'heterozygous 50%.'
2. Should the PMS2 truncating variant be reported as a germline Lynch syndrome diagnosis? Justify with the evidence available and state the next steps.
Read the worked answer
No. A tumor-only VAF of 47% in a sample with ~45% tumor content is fully compatible with a somatic event; in fact, a VAF near 50% in a tumor with roughly equal tumor/stromal admixture is the expected appearance of a heterozygous somatic variant, and it also could reflect a germline variant. Tumor-only sequencing cannot distinguish these. Additional caveats: (a) the amplicon had only 180x coverage, below the validated 500x used to support the 5% LOD, so the VAF estimate carries wider uncertainty; (b) FFPE artifacts and PMS2 pseudogene (PMS2CL) homology can generate false truncating calls; (c) absence of family history does not exclude Lynch syndrome (de novo and reduced penetrance occur). Correct action: do not issue a germline diagnosis from tumor tissue. Recommend dedicated germline testing (blood/saliva) with a validated hereditary cancer panel that includes PMS2 and addresses pseudogene co-amplification, plus genetic counseling. If germline testing is negative and clinical suspicion remains, consider tumor IHC for PMS2/MSH6 and microsatellite instability testing to assess mismatch repair deficiency.
3. The lab wants to lower its validated LOD from 5% to 2% VAF. What validation evidence would you require before accepting that claim?
Read the worked answer
Lowering the LOD is a change to a validated performance characteristic and requires revalidation. I would require: (1) a dilution-series accuracy study using cell-line or reference material with known VAFs spanning 1–10% (e.g., 1, 2, 3, 5, 10%) in replicate (typically ≥3 replicates per level, ≥20 total informative observations) to establish the LoD with ≥95% detection; (2) confirmation that the bioinformatic pipeline's variant caller, minimum read support, and strand-bias filters are tuned and validated for low-VAF calls; (3) assessment of background error rate per base and per amplicon, including FFPE-derived artifacts (C:G>T:A deamination) and oxidative damage; (4) reproducibility across operators, runs, and reagent lots; (5) defined minimum coverage (e.g., ≥1000x) and minimum alternate read count (e.g., ≥10 reads) supporting a 2% call; (6) a documented policy for reporting low-VAF calls with interpretive caveats. Only after these data demonstrate acceptable sensitivity and specificity should the LOD be changed and the report language updated.