The polite story about consumer DNA tests is that they are not for medical use. The impolite story is that the company already has the medical-grade data on you; they have decided that it is not for medical use.
Pay your two hundred dollars, send your spit, and a few weeks later the report arrives. Ancestry. Eye colour. Earwax type. Whether you can smell asparagus in your urine. Twenty-three things, give or take, all presented with the editorial register of a museum gift shop.
The raw file underneath that report has somewhere around seven hundred thousand markers in it. Single-nucleotide polymorphisms; SNPs; the small one-letter differences in DNA that the literature has spent the last twenty years arguing about. The company chose twenty-three of them, dressed them in plain English, and called it a product.
Step one: get the raw file
Every consumer testing company is required to let you download the raw .csv. They make it harder than it ought to be; three or four clicks deep, behind a 'Manage' menu, behind an 'Are you sure?' modal; but it is yours, and they have to hand it over. On MyHeritage it lives under DNA → Manage → Download Raw Data. The same dance exists on Ancestry and 23andMe, with the menu names rearranged for variety.
What lands in your downloads folder is a text file of about thirteen megabytes. Roughly seven hundred thousand rows; each row a SNP id, a chromosome number, a position, and your two-allele call. It looks like nothing in particular. It is everything they have on you.
Step two: read it against the literature
The literature has a wiki. It is called SNPedia. For every SNP that anyone has ever published about, there is a wiki page summarising what is known: which allele is associated with what, in which population, with what magnitude, with what quality of evidence.
You do not have to read SNPedia yourself. A small service called Promethease reads your raw file against SNPedia for you, in about ten minutes, and produces a PDF that is roughly forty pages long. Drug response. Cancer alleles. Lipid metabolism. Caffeine half-life. Whether you metabolise warfarin like most people. Whether the second Pfizer shot was likely to hit you harder than the first.
Five dollars. They have been doing this since 2010.
Step three: read the report with two scepticism filters on
The first filter is the magnitude system. Every finding is scored, broadly, by how meaningful the underlying association is. A magnitude-1 finding is interesting; a magnitude-4 finding wants to be brought up with a doctor. There are filtering controls on the report. Use them.
The second filter is evidence quality. The GRADE handbook, the standard reference for medical evidence rating, distinguishes between findings from single observational studies, findings replicated across cohorts, and findings confirmed by meta-analyses. SNPedia annotates this where it can. A single-study finding is a hypothesis to investigate; a meta-analysed finding is something to know.
Most of the report will be the former. That is fine. You are not making clinical decisions from a forty-page PDF; you are walking into your next physician visit knowing which questions to ask.
Why this works
The companies are not lying when they say the test is not for medical use. They mean: we have not done the clinical-validity work, we are not regulated as a diagnostic device, do not take this report into surgery. That is a reasonable defensive posture and they are correct to hold it.
What they elide is that the raw data is medically meaningful in the hands of someone who knows the literature, which is to say: in the hands of any literate adult who is willing to spend an hour reading SNPedia and forty pages of their own report. Promethease is not a regulated device either. It is a literature lookup tool. The science was done by people in white coats decades before any of these companies existed.
What the report changed for me
Six things, in roughly descending order of consequence. I will not list mine; mine are nobody's business. What I will say is that the things you find tend to be in the same shape: a drug your family has been taking that turns out to metabolise unusually in you; a screening interval that is more important to keep up than your standard insurance pamphlet suggested; a dietary fact that explains a symptom you had stopped noticing.
None of it overturned my life. All of it sharpened it. A specialist-grade report on yourself, generated from a file the company already had, for the cost of a sandwich. The marketing was the expensive part.