How to compare pet microbiome tests in Australia.

1. Does the read length support the names in the report?

Most at-home microbiome tests sequence the 16S ribosomal RNA gene, which is present in all bacteria and differs enough between species to be used for identification. The catch is that providers sequence different amounts of it.

Short-read amplicon sequencing targets one variable region, commonly V3 to V4, at roughly 460 base pairs. That length generally supports identification to the level of genus. Full-length 16S sequencing reads the whole gene at about 1,500 base pairs. Shotgun metagenomic sequencing skips the gene entirely and reads all DNA in the sample, bacterial and otherwise.

What to ask. Which region was amplified, on which platform, against which reference database, and at what confidence threshold was a species-level name assigned. If a report distinguishes Campylobacter jejuni from Campylobacter upsaliensis, or names Escherichia coli or Staphylococcus aureus, those are calls that a short amplicon cannot reliably make, and the provider should be able to say what supports them.

Our answer. BAARK runs shotgun metagenomic sequencing on Oxford Nanopore R10.4.1 chemistry for the gut sample. On our reference canine run the mean read length was 1,940 base pairs with an N50 of 3,260. We publish the platform and the reference database version, GTDB r214, and describe the classification step. The classifier is named in the technical documentation supplied with a raw data export. There is more detail in how we measure the gut microbiome.

2. Is the health score validated, and against what?

Nearly every provider returns a single headline number or grade. That number is the part of the report an owner will act on, and it is usually the part with the least evidence behind it.

A validated index states its inputs, its cut-offs, and how well it separates animals with a known diagnosis from healthy ones. The published canine benchmark is the Dysbiosis Index described by AlShawaqfeh and colleagues in 2017, built on eight quantitative PCR assays, which reported an area under the curve of 0.93 and 74 per cent sensitivity with 95 per cent specificity at a threshold of zero. Those are the kind of figures a score should come with.

What to ask. What goes into the score, where the cut-offs sit, how many animals it was tested on, and what its sensitivity and specificity are. A ladder running from critical to excellent, with no stated cut-offs and no performance figures, has not been shown to separate anything. It may still be a reasonable summary. It is not a validated index, and it should not be described as one.

Our answer. BAARK reports the published Dysbiosis Index using its eight quantitative PCR assays, and cites the figures above rather than substituting our own. Where we report anything beyond that index, we label it as our own summary and not as a validated result.

3. Are the numbers relative or absolute?

This is the distinction that causes the most confusion, and it changes what a result means.

Sequencing returns proportions of a fixed pool of reads. If one organism blooms, every other organism's percentage falls even though its actual quantity in the animal has not changed at all. Quantitative PCR measures each chosen target against a standard curve and returns copies per gram, a number that does not move because something else moved.

Clostridium hiranonis is the clearest case. It converts primary bile acids to secondary bile acids, and its loss is one of the better established signals of canine dysbiosis. If a report shows it at 0.0 per cent relative abundance, that reading cannot tell you whether the organism is genuinely gone or simply crowded out of the read pool by an overgrowth of something else. Those two situations call for different responses.

What to ask. For each number in the report, is this a share of reads or a measured quantity. If your pet is retested in three months, will the two results be directly comparable.

Our answer. BAARK reports sequencing as relative abundance and labels it that way, and reports the quantitative PCR targets as absolute quantities against a standard curve. We do not present one as the other.

4. Are parasites included, and by what method?

The 16S ribosomal RNA gene is a bacterial gene. Parasites and worms are eukaryotes and do not carry it. A test built on 16S sequencing cannot detect Giardia, hookworm, roundworm or whipworm at any sequencing depth, because the target simply is not there to find.

Parasite coverage therefore requires either shotgun metagenomic sequencing, which reads all DNA present regardless of organism, or a targeted quantitative PCR panel against named parasite targets. Both are legitimate. A 16S-only test that implies parasite coverage is not.

What to ask. Is a parasite result included, which organisms are on the list, and by which method.

Our answer. BAARK reads parasites from the shotgun run and additionally runs a targeted quantitative PCR panel of seven parasite and worm targets, because low-abundance organisms can fall below the read threshold in sequencing. The full target list is on what we test.

5. Does the sample report hold together?

Ask any provider for a full sample report before you buy. It takes a few minutes to check and it tells you more than the marketing does. Five things to look at.

• Do the percentages add up. If a top-fifteen table sums to 82 per cent and there is no row labelled other or unclassified, roughly a fifth of the community is unaccounted for and the report is not saying where it went.
• Do the categories overlap. If a report presents core, beneficial and harmful groups as separate categories, the same organism should not appear in two of them with the same value.
• Do the reference ranges have width. A healthy range printed as 0.0 to 0.0 is not a reference range, and it implies a limit of detection of exactly zero, which no assay has.
• Is absence reported as normal. An organism at 0.00 per cent marked normal, purely because the bottom of its reference range happens to be zero, is a scoring rule rather than a finding.
• Is the taxonomy consistent. A table headed species should not contain family names. Lachnospiraceae, for instance, is a family, not a species.

6. Is there a quality control result in the report?

Sequencing runs vary. Samples degrade, extractions fail, and a run can produce too few reads to support the conclusions drawn from it. None of that is unusual. What matters is whether the provider tells you when it happens.

What to ask. Does the report state how many reads the sample produced and whether that met a stated minimum. Were negative and positive controls run alongside the sample. What happens if a sample fails, and is a repeat test charged for.

Our answer. BAARK prints the read count, mean read length and N50 for your pet's run on the report itself. Our sample reports for a dog and a cat show exactly what that looks like.

7. Where is the laboratory, and how long is the sample in transit?

Faecal microbial communities begin to shift as soon as the sample leaves the animal. For a mail-in test, the stabilising buffer and the transit time can matter more to the result than the choice of sequencing platform. A profile dominated by one or two organisms with the usual community members near zero is a pattern that can arise from a genuine biological state, and can also arise from a sample that spent too long getting to the laboratory. A good provider distinguishes the two and says so on the report.

What to ask. Which stabiliser is used and how many days it is validated for at room temperature. Where is the laboratory. Is the sample sequenced in Australia or sent overseas.

Our answer. BAARK sequences in its own Melbourne laboratory. The stabiliser holds samples at room temperature for up to 14 days in transit, and no ice or special handling is needed.

8. Is the advertised price the whole price?

Gut and oral testing are often sold as separate products, with separate kits, separate collection and separate return postage. The advertised price of one is not the cost of both, and the difference can be substantial.

What to ask. Work out the total for the samples you actually want rather than comparing headline prices. Then check, separately: is return postage included, is a repeat test charged for if the first sample fails, is a veterinary consultation included or extra, is the price a one-off or a subscription, and is it quoted in Australian dollars.

Our answer. BAARK's introductory price is $249 for gut and oral together from a single kit, in Australian dollars, with free return postage and no subscription. It is an introductory price and it will change. Current pricing is on the order page.

9. Can you get your raw data?

The sequencing reads and the full species abundance table are the evidence the report is built on. A provider willing to release them is one whose result can be checked by someone else, including your veterinarian or a second laboratory. A provider unwilling to release them is asking you to take the report on trust.

What to ask. Can I have the raw reads and the full abundance table, in what format, at what cost, and how long are they retained.

Our answer. Yes, on request and at no charge. The terms are set out in our raw data access policy.

A one-page summary of our own answers

So that this page can be used against us as easily as against anyone else, here is BAARK on every question above.

BAARK's answer to each of the nine comparison questions

Question	BAARK
Method and read length	Shotgun metagenomic sequencing, Oxford Nanopore R10.4.1. Reference run mean read length 1,940 bp, N50 3,260 bp.
Validated score	Published Dysbiosis Index, 8 qPCR assays, AUC 0.93, 74% sensitivity and 95% specificity at threshold 0 (AlShawaqfeh et al. 2017).
Relative or absolute	Both, labelled separately. Sequencing reported as relative abundance; qPCR targets as copies per gram against a standard curve.
Parasites	Read from the shotgun run, plus a targeted qPCR panel of 7 parasite and worm targets.
Oral sample	Included. Targeted canine oral qPCR panel from the same kit, no separate purchase.
Quality control on the report	Read count, mean read length and N50 printed for every run.
Laboratory	Melbourne, Australia. Stabiliser validated to 14 days at room temperature.
Price	$249 AUD introductory, gut and oral together, one kit, free return postage, no subscription. Introductory and subject to change.
Turnaround	5 to 7 business days from sample receipt at our Melbourne laboratory.
Raw data	Released on request at no charge. See the raw data access policy.

What we are not claiming

BAARK is a wellness screening service, not a diagnostic test, and it does not prevent disease. A microbiome result describes what is in a sample at the time it was taken. It does not diagnose a condition, and it is not a substitute for examination by a veterinarian. If your animal is unwell, see a vet first and bring the report with you.

Read further

• How we measure the gut microbiome, on sequencing method and why read length matters
• Dysbiosis and digestive disease in dogs, on what the Dysbiosis Index measures
• The oral microbiome in dogs and cats, on why the oral panel is quantitative PCR rather than sequencing
• What we test, the full target lists
• Raw data access policy