All articles

Compound heterozygosity needs phasing.

Two heterozygous variants in a recessive gene explain a case only if they sit on different copies. Without phase, the report is guessing.

A woman of thirty-one with progressive sensorineural hearing loss and retinitis pigmentosa has a singleton exome. It finds two heterozygous variants in USH2A: one well established as pathogenic, one missense change of uncertain significance. The report describes a likely compound heterozygote, and the clinic tells her that the cause has been found. She was adopted, and no parental sample can be obtained.

Three years later a long-read genome, done for a research study, shows both variants on the same copy of the chromosome. The gene was still the best candidate for her condition, and the second allele had never been looked for, because the case had been closed.

What the field knows

Phase is which variants sit together on the same copy of a chromosome. Tewhey and colleagues (2011) set out why it matters in medical genomics: the effect of two variants in one gene depends on whether they disrupt the same copy, leaving the other intact, or different copies, leaving none. Short-read sequencing calls each position on its own and, unless two variants fall within one read or one read pair, says nothing about which copy each is on.

Phasing methods fall into families. Read-based phasing links variants that occur within the same fragment; HapCUT2 (Edge, Bafna and Bansal, 2017) assembles haplotypes from such links for any sequencing technology, and does well with long or linked reads and poorly with standard short reads where the variants are far apart. Statistical phasing (Browning and Browning, 2011) uses haplotype frequencies from a reference panel, which works for common variants and is unreliable for the rare ones that cause Mendelian disease. Pedigree phasing uses the parents: if one variant came from each, they are on different copies. Choi and colleagues (2018) compared the strategies on whole genomes and found that for rare variants more than a read length apart, only long-range data or parental genotypes gave a dependable answer.

The interpretation standards assume the answer exists. The ACMG and AMP framework (Richards et al., 2015) treats detection in trans with a pathogenic variant as moderate evidence that a second variant in a recessive gene is pathogenic, and states that this needs the phase to be confirmed, by testing the parents or by other means. Where phase is unknown, that line of evidence is unavailable, and a variant of uncertain significance stays uncertain.

The question arises often. In outbred populations, homozygosity for the same rare allele is uncommon, so most recessive diagnoses rest on two different heterozygous variants, and every one of them depends on phase. Whenever the parents have not been tested and the variants are too far apart to share a read, the laboratory is relying on an assumption where the framework asks for a measurement.

Where practice breaks

The phrase in the report does the damage. 'Two heterozygous variants were identified in this gene, consistent with compound heterozygosity' is read by a clinician as the cause having been found. It was written by a scientist who meant that if the variants are in trans this is the answer, and that the laboratory cannot tell. The conditional is in the scientist's head and absent from the page.

Parents are missing more often than the textbooks assume: adopted patients, estranged families, deceased parents, a parent abroad, a family that declines. A laboratory whose only phasing route is parental testing has no route for these patients, and the case is either over-called or parked. Singleton testing is also increasingly the default on cost grounds, which moves phasing from something a trio gives for free to something that has to be asked for, paid for and remembered.

When phase is unknown the interpretation is usually settled by plausibility. The two variants are rare, the gene fits the phenotype, the first variant is known to be pathogenic; it feels unlikely that the second should also sit on the same copy. Sometimes it does. Variants in cis on a shared founder haplotype are well described, and a patient with one pathogenic allele and a second rare variant beside it on the same copy has an undiscovered second allele, which may be a deletion, a deep intronic change or a structural variant the exome cannot see. Calling the case solved ends that search.

The uncertainty also fails to survive into the record. Years later the case reads as solved. If a relative asks for carrier testing, or a therapy trial requires confirmed biallelic variants, the unknown phase has to be rediscovered from a report that never mentioned it.

How Tulparus handles it

The workbench treats phase as evidence with its own state, separate from the variants. For every pair of heterozygous variants in a recessive candidate, the case records whether phase is established in trans, established in cis or unknown, and how it was established: parental genotypes, read-based linkage where the pipeline could provide it, or not at all. The hypothesis that the gene explains the case carries the gap as part of its structure, and the comparison with other hypotheses shows it.

Where phase is unknown, the missing evidence is named and the next step is proposed with what it would take: parental samples if they can be obtained, a targeted long-read or allele-specific assay if they cannot, and what each result would do to the hypothesis. The step sits in the case plan beside the other open questions, and the case stays open until someone decides otherwise.

The report is rendered from the structured case, so it says what the case knows. If phase is unknown, the report says that two heterozygous variants were found, that their phase has not been established, and which test would establish it. A clinician cannot read 'consistent with' as confirmed because the sentence is not there. When phase is later established, the case is versioned and the report is regenerated, and the earlier version is kept with its date.

Phase can arrive late and from an unexpected direction. If a relative is tested years afterwards, their genotype at the two positions may settle the question, and the workbench records the inference with its source when the new sample is added to the family. A case that was closed on an assumption is reopened by the fact, and the clinician who signed the earlier report is told.

For the woman above, a case written this way would have stayed open: one established pathogenic allele, one variant of uncertain significance, a named gap and a proposed assay. The search for the second allele would have continued.

What you can do tomorrow

The change is mostly in what gets written down.

  • Read your report template. If it contains 'consistent with compound heterozygosity' without a phase statement, add one: in trans and how, in cis, or not established.
  • List the recessive diagnoses your laboratory reported in the last two years and check which had parents tested. Any without is a case with unknown phase.
  • Find out which read-based phasing your pipeline can already do. Two variants within a read pair can be phased from the alignment you already hold.
  • When parents are unavailable, write the long-read or allele-specific assay into the plan rather than closing the case.

Sources

  1. Tewhey R, Bansal V, Torkamani A, Topol EJ, Schork NJ. The importance of phase information for human genomics. Nature Reviews Genetics, 2011.
  2. Edge P, Bafna V, Bansal V. HapCUT2: robust and accurate haplotype assembly for diverse sequencing technologies. Genome Research, 2017.
  3. Browning SR, Browning BL. Haplotype phasing: existing methods and new developments. Nature Reviews Genetics, 2011.
  4. Choi Y, Chan AP, Kirkness E, Telenti A, Schork NJ. Comparison of phasing strategies for whole human genomes. PLoS Genetics, 2018.
  5. Richards S, Aziz N, Bale S, et al. Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genetics in Medicine, 2015.

Two variants, one gene.See whether the phase is known.

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