Companion animal genomics and diagnostics

Dogs and cats are exceptionally good subjects for genetics, and for an uncomfortable reason. Intense selection and closed studbooks produced breeds that are, in population terms, extended families: long haplotype blocks, low within-breed diversity, and disease alleles concentrated at frequencies that would be vanishingly rare in an outbred population. This makes mapping easy — a handful of animals per breed can localise a Mendelian trait that would need thousands of humans — and it makes the results narrow in exactly the same proportion.

Why the reference panel decides the answer

A variant is discovered in the breed where the disease clusters, on that breed’s haplotype background. Whether the tested marker is the causal change or merely a marker linked to it within that breed’s shared segment is often not resolved at discovery. Outside the breed the linkage need not hold: the same marker may travel with a normal allele, so a positive result in another breed or a mixed-breed dog can mean nothing at all. The converse also occurs, since a different causal variant in the same gene will be missed by a test that types only one position.

Direct-to-consumer panels compound this by typing hundreds of variants on every sample regardless of breed. The genotype is usually accurate — array genotyping of a known site is a reliable measurement. The interpretation is what fails, and it fails predictably: a variant whose association is established in one breed is reported as a risk in dogs where its penetrance has never been measured. Reported cases of owners considering euthanasia on the strength of such a result have been raised in the veterinary literature, and the recurring recommendation is confirmatory testing and clinical correlation before any decision.

Breed ancestry estimation has a separate limitation. It is a statistical assignment against a reference set of pedigree animals, so ancestry can only be reported in terms of the breeds in that reference; village dogs and regional landraces are represented poorly or not at all, and small percentages in a report are the least reliable part of it.

Two kinds of test, often sold as one

The useful distinction is not technical but clinical. A Mendelian test for a fully penetrant variant with a known consequence changes decisions: the MDR1 variant affecting drug transport tells a veterinarian which drugs and doses are dangerous in that animal; a progressive retinal atrophy or a coagulopathy variant informs breeding; a known cardiomyopathy variant in a breed where it is validated informs screening. The result is actionable because penetrance and consequence are established in the population being tested.

A risk panel reporting incompletely penetrant associations of small effect is a different object. Without breed-specific penetrance estimates it cannot give a probability, and without a probability there is no decision to change. It can still cause harm through anxiety, unnecessary intervention or breeding exclusion that narrows an already narrow gene pool.

The direction of the field is towards whole-genome sequencing, which removes the fixed-marker problem but not the interpretation problem: it will find more variants of unknown significance, and knowing what a variant does in a breed still requires phenotyped animals from that breed.