What a Pet DNA Test Actually Tells You

What a Pet DNA Test Actually Tells You

A pet DNA panel is not one test. It is dozens of them bundled together, and their quality varies enormously. Some name a mutation that reliably causes a disease, validated in the breed you are testing. Others report a change that has been catalogued but never shown to cause anything at all.

They arrive on the same page, at the same price, formatted identically, with nothing telling you which is which.

This is a guide to reading the report.

First, the Decoder

Almost every variant name is a compressed sentence. Once you can read it, the reports stop being intimidating.

The single letters are amino acids, the building blocks proteins are made from. A is alanine, P is proline, R is arginine, W is tryptophan, C is cysteine, G is glycine. There are twenty in total and you never need to memorize them: the point of the notation is the position and the fact that something changed, not the chemistry.

Then, the Four Words That Do All the Work

Penetrance is the one that matters most. Of all the animals carrying a risk genotype, what proportion ever actually develop the condition? Some variants are fully penetrant, so the genotype is effectively a diagnosis. Others are barely penetrant, and the genotype means almost nothing.

Expressivity is the follow-on question. Among those that do develop it, how badly?

Genotype is what the DNA says. Phenotype is what the animal actually is. When they disagree, the animal is not wrong.

And one number: an odds ratio compares the chance of disease in animals with the genotype against animals without it. An odds ratio of 1.0 means no difference at all. Read it alongside its confidence interval, because if that range crosses 1.0, the finding is not statistically significant no matter what the headline number looks like.

The Worked Example That Explains Everything

The Maine Coon heart variant, A31P in the MYBPC3 gene, is the most instructive result in companion animal genetics.

An odds ratio of 1.0 for the one-copy cats. Not slightly raised. Not a mild concern. Literally no detectable difference from cats with no mutation at all, in a cohort of 332 cats. The study's authors concluded the mutation must have very low penetrance in that group.

Now the other direction. Ten of the 21 cats that actually had heart disease in that cohort carried no mutation whatsoever. The genotype missed half the phenotype.

So a positive one-copy result on this variant is not a diagnosis, and a clear result is not a clearance. The veterinary consensus statement says exactly that, and rates the evidence high: Maine Coons and Ragdolls testing negative for these variants have still been diagnosed with the disease, so echocardiographic screening should continue anyway.

Where the Tests Are Genuinely Good

It would be wrong to leave you thinking these tests are all oversold. They are not.

A large study of over a million dogs found complete penetrance, 100%, for ten variants including canine leukocyte adhesion deficiency type III, cone-rod dystrophy 2, and hemophilia A. Tested in the breed it was validated in, a well-characterized recessive variant does exactly what it says.

The other genuinely good news is what testing has achieved. Feline polycystic kidney disease once ran at roughly 36% to 50% of Persians. Where testing has been systematic, carrier frequency fell 42.6% across 14 breeds in three years. A study of 5,512 Pembroke Welsh Corgis found degenerative myelopathy homozygotes dropping from 14.5% to 2.9% between 2019 and 2022, with no loss of genetic diversity at all and effective population size actually rising, because breeders used the test to widen their search rather than narrow it.

Used properly, these tests work.

Where They Do Not

Variants of unknown significance, sold anyway

Six feline heart-disease variants are sold commercially. When reviewers applied formal classification criteria across 21,772 cats, two were pathogenic and three were variants of unknown significance. Their recommendation was to limit routine testing to two variants in two breeds and to avoid screening for the three unknowns entirely, for lack of evidence they cause disease.

The clearest case is ALMS1, marketed as a Sphynx heart test. In a New Zealand cohort it was present in 70.9% of cats and showed no association whatsoever with the diagnosis. That same cohort still reached 40% heart disease prevalence. The test finds something. It does not find the disease.

Variants sold outside the breed they were validated in

This is the best-evidenced problem of the lot, and the sentence to remember comes from the researchers themselves: discovery of a disease-implicated variant on another genetic background does not equal confirmation that presence of the variant leads to disease onset.

In a study of over a million dogs, 26 disease variants turned up in 65 additional purebred breeds with no prior characterization in the literature. Named examples where the variant is probably not doing what the panel implies include a cystinuria variant found at 12% to 26% in Golden Retrievers and Swedish Vallhunds, and two dilated cardiomyopathy variants discovered in Dobermans and now found across many breeds.

The Dachshund retinal variant is the cleanest illustration. The OFA states that over 25 breeds carry it, including the Beagle, Rottweiler and French Bulldog, but that outside Miniature Longhaired Dachshunds and English Springer Spaniels the risk of actually developing disease is low.

Linked marker tests, which are not the same thing

A causal-variant test looks at the actual disease-causing change. A linked marker test looks at a signpost sitting nearby, used when the real mutation is unknown. It works only while the two stay together on the chromosome, and it fails in lines where they have separated.

The Bedlington Terrier copper toxicosis test is the textbook failure. The marker sat about 13,500 DNA letters away from the actual mutation, and a chromosome was eventually found carrying the disease-associated marker alongside a perfectly healthy gene, producing false positives.

How to tell which you are buying: ask the lab to name the variant in full notation, gene plus position plus change. A causal test can always answer. A linked marker test cannot. Watch for the words marker, linked, indirect, haplotype and associated.

And one test that simply did not work

A 17-marker hip dysplasia test, patented and sold commercially, was put through independent prospective validation in 935 German Shepherds followed for three years. Its area under the curve was 0.523, where 0.5 is a coin toss. The odds ratio pointed the wrong way and was statistically indistinguishable from noise. The authors' conclusion was that the test is unsuitable for individual risk assessment.

Nobody Is Checking

This is the part most owners assume is handled, and it is not. From the standards paper written by the field itself:

"Currently, there is no regulatory oversight or testing standards for diagnostic veterinary genetic testing laboratories."

Veterinary genetic labs are not required to be accredited. No external proficiency program exists for disease-variant genotyping. The biennial ISAG comparison test that labs sometimes cite is explicit that it is not a proficiency or accreditation test, and in any case it covers parentage and identity markers, not disease variants. A lab can perform perfectly in it while genotyping disease variants badly.

What this looks like in practice: researchers sent samples from 12 registered purebred dogs to six direct-to-consumer companies, swapping the submitted photographs on half of them. Two companies matched the registered breed on every dog. One matched zero of nine, with results that tracked the photograph rather than the DNA. One company reported a purebred Bulldog as 10% to 20% wolf. Four of the twelve dogs were assigned ancestry banned by more than half of home insurance companies, despite none actually being such a breed.

That study tested breed ancestry, not disease variants, and the distinction matters. But it tells you what "no oversight" means concretely.

What to Actually Do With a Result

Never breeding carrier to carrier is close to a free intervention: it removes affected offspring without removing a single animal from the gene pool.

The opposite mistake is real too. The OFA warns in its own guidance that with 29.1% carriers and 14.8% at-risk dogs, an overly aggressive program to eliminate every non-clear German Shepherd would remove roughly 44% of tested dogs and could be devastating to the breed. The failure mode there is not the test. It is replacing a broad cull with a narrow set of popular clear sires, and the Corgi study above shows it is entirely avoidable.

There is a subtler trap as well. Selecting hard against one trait drags correlated traits with it. In 5,038 Maine Coons screened for hip dysplasia over 20 years, severity was genetically correlated with body mass, so selection reduced dysplasia and made the cats about 33 grams smaller each generation. The same mechanism ran in reverse in the Labrador: selecting for the food motivation that makes a good assistance dog pushed the appetite variant to a 45% allele frequency in assistance breeding stock against 12% in the general population.

What Is Worth Buying for a Pet

Worth it, because the result changes something tomorrow:

• MDR1, unambiguously. A causal variant affecting drug safety, relevant to any herding breed or herding cross. Rough Collies run about 70%, Australian Shepherds 50%, mixed breeds 5%. Your vet needs to know before prescribing ivermectin-class wormers, high-dose loperamide or certain sedatives.
• Other pharmacogenetic and acute-risk variants, such as malignant hyperthermia and clotting disorders, which change anesthesia and surgical planning.
• A validated causal variant for a manageable disease, where early detection changes care.

Not worth it for a pet:

• Late-onset risk variants with no available intervention. An at-risk degenerative myelopathy result on a healthy three-year-old changes nothing you can act on, and its penetrance is strongly age-dependent: in one survey no at-risk dog under 10 had signs, while all seven over 10 did.
• Anything classed a variant of unknown significance.
• Breed ancestry, if you intend to act on it.
• Broad panels bought for reassurance. More tests are not necessarily better.

For a breeding animal, the breed-relevant validated variants, ideally from a lab invested in that specific disease, plus a diversity measure alongside the disease panel. The average purebred dog across 227 breeds already carries an inbreeding coefficient of about 0.249, which is roughly what you would get from mating full siblings, and the more inbred breeds carry measurably higher morbidity.

The Honest Summary

A DNA test tells you what an animal carries. A clinical examination tells you what it has. Neither on its own tells you what it will develop or pass on.

Buy the tests that change a decision. Ignore the ones that do not. And when a result and the animal in front of you disagree, believe the animal.

Sources & Further Reading