Allele balance, maternal mosaicism and recurrence risk in a trio exome
A 3-year-old boy with epileptic encephalopathy undergoes clinical trio whole-exome sequencing (mean on-target coverage 92x). A heterozygous nonsense variant in a known disease gene is reported in the proband with 26 alternate reads out of 84 (VAF 0.31). Sanger sequencing of the proband confirms the variant, but the peak heights are visibly unequal. Parental blood samples, sequenced to 32x and 28x respectively, show no alternate reads. Both parents are clinically normal and there is no family history. Because the parents are planning another pregnancy, the laboratory re-contacts them; a maternal saliva sample, collected later, shows 3 alternate reads out of 45 (VAF 0.067). The current report describes the variant as 'de novo, autosomal dominant, recurrence risk negligible'. The mother asks what the chance is that a future child is affected.
1. Is a VAF of 0.31 at 84x coverage consistent with a constitutional heterozygous variant? What alternatives must be excluded before the variant is called?
Read the worked answer
Test it. With n = 84 and p = 0.5, the expected alternate count is 42 and the standard deviation is √(84 × 0.25) = 4.58. The observed 26 gives z = (26 − 42)/4.58 = −3.5, a two-sided probability of roughly 4 × 10⁻⁴. That is unlikely to arise by chance from a clean heterozygous site, and the unequal Sanger peak heights point the same way.
Alternatives to exclude:
- Somatic mosaicism in the child - the variant arose post-zygotically, so only a fraction of blood cells carry it. Buccal or fibroblast testing would show a different VAF if so.
- Reference-allele or mapping bias - a variant in a region with a paralogous pseudogene or a repetitive context can be systematically under-called. Check MAPQ, strand bias and read position, and inspect the locus manually.
- Copy-number complexity at the locus - a duplication of the reference allele, or a deletion in cis, distorts the expected 1:1 ratio. Run a read-depth/CNV caller on the exome and confirm by an orthogonal method.
- Sample mix-up or contamination - check the sex karyotype from the data, contamination estimates and the sample chain of custody.
Do not sign out 'heterozygous, de novo' until the VAF anomaly is explained. Confirm with an amplicon-based deep-sequencing assay or droplet digital PCR, which quantify allele fraction directly rather than relying on capture-based exome read counts.
2. What recurrence risk would you quote the mother, and how should the report be amended?
Read the worked answer
The maternal saliva result changes the answer. Three alternate reads out of 45 (6.7%) is above the background error rate of a well-run amplicon or exome pipeline and indicates maternal somatic mosaicism.
What can and cannot be said:
- Gonadal involvement cannot be measured directly. Saliva derives from ectoderm, blood from mesoderm; detecting the variant in a second germ layer makes pre-meiotic mosaicism likely and gonadal involvement plausible, but the gonadal fraction may be anywhere from very low to near 50%.
- Published empirical recurrence risks for confirmed parental mosaicism cluster in the 5–15% range rather than the 25% Mendelian figure, and can be higher when the mosaic fraction is large.
- It is not negligible, and 'recurrence risk negligible' is an incorrect statement in the report.
Quotation to give the family: 'roughly a 5–10% chance, considerably higher than the general population risk but well below the 25% seen with a carrier parent'.
The report should be amended to state that the variant is de novo in the proband with evidence of maternal somatic mosaicism at ~7% in saliva, that recurrence risk is elevated and not negligible, and that prenatal diagnosis or preimplantation genetic testing is available and technically feasible. A genetic counsellor should deliver this, and the amended report should be reissued with a corrected classification line rather than an addendum that leaves the original wording standing.
3. Why did the parental blood testing at ~30x coverage miss a variant present at ~7%?
Read the worked answer
Because a negative result at low coverage is not the same as absence.
Quantify it. If the mother's blood were mosaic at 6.7%, the probability of observing zero alternate reads at 28x is (1 − 0.067)^28 = 0.933^28 ≈ 0.14. So roughly a one-in-seven chance of a false negative even if the variant is genuinely present in blood at that level. The expected count would be under two reads, which most pipelines filter out as noise anyway.
As a rule of thumb, the 95% detection limit of a variant by read counting is about three alternate reads, so at 30x the limit is roughly 10%; at 200x it drops to about 2%; at 500x, below 1%. A 'negative' parental blood at 30x therefore excludes only mosaicism above about 10%.
Practical recommendations:
- Re-test both parents' blood with a targeted deep amplicon assay at ≥200x, plus the mother's saliva and, if accessible, a skin or buccal sample.
- Remember that absence in blood does not exclude germline mosaicism confined to the gonad; a negative blood result in a family with one affected child still leaves a small but real recurrence risk that should be quoted.
- For the family's next pregnancy, the practical route is prenatal testing or preimplantation genetic testing using the variant itself, which sidesteps the uncertainty in the recurrence figure entirely.