All articles

Beyond the small variant.

An exomeSequencing of the protein-coding parts of the genome, about one to two percent of it, where most known disease variants sit. pipeline tuned for small variants is blind to deletions, duplications and repeat expansions. Some negative cases were never looked at properly.

A girl of six with a recessive metabolic phenotype, biochemically clear, has a gene panel and then an exome. Both find one heterozygous pathogenic variant in the expected gene and nothing else. Her parents are both carriers, of the same variant. The case is reported as a carrier with an unexplained phenotype and sits for two years.

A genome, done eventually under a research protocol, shows a deletion of two exons on the other allele. It had been in the exome's coverage data all along, as a run of exons at half depth, in a file nobody opened because the pipeline reported only small variants.

What the field knows

Short-read pipelines were built around the variants short reads see well: single-nucleotide changes and insertions or deletions of a few bases, called from the reads that span them. Larger changes leave different traces. A deletion lowers the depth of coverage over the region it removes; a duplication raises it; a breakpoint produces reads that are split or whose mates map far apart; a repeat expansion produces reads that are entirely repeat and cannot be placed. Each trace needs its own caller.

The callers exist and have been evaluated in clinical settings. Depth-based methods for exome data, such as XHMM (Fromer et al., 2012), find multi-exon deletions and duplications from the same reads the small-variant caller used. Gross and colleagues (2019) described deploying copy-number calling within a clinical genome service and the diagnoses it added, and Riggs and colleagues (2020) published the technical standards that let a copy-number variant be classified with the same discipline as a sequence variant. For repeat expansions, ExpansionHunter (Dolzhenko et al., 2017) genotypes known loci from genome data, and Ibañez and colleagues (2022) showed in a national service that genome sequencing with repeat calling could detect the expansions behind common neurological disorders with accuracy fit for diagnostic use.

The contribution to yield is consistent across series. Stranneheim and colleagues (2021), reporting on several thousand rare disease patients sequenced in a healthcare setting, found structural and copy-number variants and repeat expansions among the diagnoses that small-variant analysis alone would have missed. The share differs by indication, and in every indication it is present.

None of these callers is as dependable as a small-variant caller, and the field is clear about it. Exome depth varies with capture efficiency and GC content as well as with copy number, so a depth-based call needs a well-matched baseline and, for reporting, confirmation by an independent method such as multiplex ligation-dependent probe amplification or an array. Repeat genotyping from short reads sizes an expansion only up to the read length and reports larger ones as a lower bound. The callers find candidates. The confirmation is still part of the work.

The recessive case is where the gap is sharpest. One pathogenic small variant in the right gene, in a patient whose phenotype matches, is a strong hint that the second allele exists and was missed. A deletion of one or more exons on the other copy is the first place to look, and the exome's own coverage can often show it.

Where practice breaks

The exome pipeline reports what it was configured to report. Copy-number calling from exome depth is noisy and needs a batch of similar samples as a baseline, so many laboratories never turned it on, or turned it on for a validated gene list only. The coverage data are generated and discarded, or kept and never looked at.

The one-variant recessive case is handled by a form of words. 'Heterozygous carrier; a second variant was not identified' closes the case while sounding open. The clinician reads a carrier result and moves on. Nobody is assigned the question of where the second allele is, and the exon-level deletion that would answer it is not on anybody's list.

Repeat expansions are missed differently. They are not looked for because the phenotype did not suggest a repeat disorder to the person ordering, and the pipeline does not look for what it was not asked to. Ibañez and colleagues found expansions in patients who had been referred under other diagnoses. A pipeline that genotypes the known loci on every sample removes the need to have thought of it.

When a structural variant is found, its evidence is handled worse than a sequence variant's. The classification standards for copy-number variants (Riggs et al., 2020) are newer and less often built into tooling, so the finding arrives at the scientist as a coordinate range and a copy number, without the gene content, the dosage sensitivity of those genes or the population frequency of similar events laid out. It takes longer to assess, and so it is assessed less often.

How Tulparus handles it

The workbench treats variant classes as channels that each case is expected to have, and records which ones it does have. A case whose pipeline produced small variants and no copy-number or repeat calls is marked as such, and the gap appears on the case rather than being inferred from a silence. For an exome, the workbench reads the per-exon coverage where it is supplied and flags runs of exons at half or double the expected depth in genes relevant to the phenotype, as a prompt for a proper copy-number analysis.

The one-variant recessive hypothesis is modelled as what it is: one allele established, the second missing. The missing second allele is a named gap with its own list of places to look, in order: exon-level deletion from coverage, a deep intronic or splice region variant in the same gene, a structural rearrangement, and, if the gene has a known repeat locus, an expansion. Each step carries the data it needs and whether the case already holds them.

Copy-number and structural findings, when they arrive, are presented with the evidence the standards ask for: the genes in the interval, their dosage-sensitivity curation, the frequency of overlapping events in population data, and the phenotype match of the genes affected. The scientist classifies with the same structure used for sequence variants, and the classification is stored with its provenance.

For the girl above, the case would have shown one established allele, one named gap and, from the exome's own coverage, a flagged run of two exons at half depth in the gene concerned, with a copy-number confirmation as the first proposed step. Two years would have been a fortnight.

What you can do tomorrow

Some of this is pipeline work. The first items are not.

  • List your recessive cases closed with one pathogenic variant and a matching phenotype. Each is a second-allele search that was never run.
  • Find out whether your exome pipeline keeps per-exon coverage. If it does, a half-depth run across consecutive exons in the gene of interest can be checked by eye today.
  • Ask your pipeline team which variant classes are called on every sample and which only on request. Write the answer into the report template, so that a negative says what was and was not looked for.
  • If you sequence genomes, run repeat genotyping at the known loci on every sample rather than on suspicion.

Sources

  1. Fromer M, Moran JL, Chambert K, et al. Discovery and statistical genotyping of copy-number variation from whole-exome sequencing depth. American Journal of Human Genetics, 2012.
  2. Gross AM, Ajay SS, Rajan V, et al. Copy-number variants in clinical genome sequencing: deployment and interpretation for rare and undiagnosed disease. Genetics in Medicine, 2019.
  3. Riggs ER, Andersen EF, Cherry AM, et al. Technical standards for the interpretation and reporting of constitutional copy-number variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics (ACMG) and the Clinical Genome Resource (ClinGen). Genetics in Medicine, 2020.
  4. Dolzhenko E, van Vugt JJFA, Shaw RJ, et al. Detection of long repeat expansions from PCR-free whole-genome sequence data. Genome Research, 2017.
  5. Ibañez K, Polke J, Hagelstrom RT, et al. Whole genome sequencing for the diagnosis of neurological repeat expansion disorders in the UK: a retrospective diagnostic accuracy and prospective clinical validation study. Lancet Neurology, 2022.
  6. Stranneheim H, Lagerstedt-Robinson K, Magnusson M, et al. Integration of whole genome sequencing into a healthcare setting: high diagnostic rates across multiple clinical entities in 3219 rare disease patients. Genome Medicine, 2021.

One allele found.See where the second one is looked for.

A demo takes 20 minutes, screen shared, on synthetic cases. Bring the question from one of your own.