How we turn published research into your report
There is no way to check whether an interpretation is any good without knowing how it was produced. This page is that account: where the underlying research comes from, what we do when it disagrees with itself, and the limits we cannot engineer our way past.
How does DNA Checker turn published research into a report?
Every reading starts from a published, peer-reviewed finding and a marker that consumer arrays actually carry. We record the source, the variant, and what the research supports. Your file is then read for that variant, and the report states what the literature says about it — never a fact about you.
The order matters. Starting from the research and asking which findings a consumer file can support produces a short, defensible list. Starting from the file and asking what could plausibly be said about each marker produces a long and much more impressive list, most of which would not survive contact with the evidence.
That is why the catalogue is the size it is. It is bounded by what published work supports and what a genotyping array can actually see.
Which databases does DNA Checker use?
Six, all open and publicly funded: ClinVar and gnomAD for variant significance and population frequency, the GWAS Catalog and PGS Catalog for association studies and scores, and CPIC and PharmGKB for pharmacogenomics. Each is maintained by a public research body, and each is listed below with what we use it for.
| Source | Maintained by | What we use it for |
|---|---|---|
| ClinVar | US National Center for Biotechnology Information (NIH) | Clinical significance of individual variants, and how strongly that interpretation is supported. |
| gnomAD | Broad Institute of MIT and Harvard | How common a variant is, broken down by population — the check that stops a common variant being read as remarkable. |
| GWAS Catalog | EMBL-EBI and the US National Human Genome Research Institute | Published associations between variants and traits, with the study behind each one. |
| PGS Catalog | EMBL-EBI | Published polygenic scores, including which populations each was developed and evaluated in. |
| CPIC | Clinical Pharmacogenetics Implementation Consortium, hosted on ClinPGx (NIH-funded) | Peer-reviewed guidelines on how gene variants relate to specific medicines. |
| PharmGKB | ClinPGx, Stanford University (NIH-funded) | Curated evidence linking genetic variation to drug response, graded by strength. |
None of these is proprietary and none is ours. You can look up any marker in a report yourself, which is the point — an interpretation you cannot check is a claim you have to take on faith.
CPIC and PharmGKB have since been consolidated into a single site, ClinPGx. Both bodies of work continue there, and the links above go to their current homes rather than the older addresses.
What does "not covered" mean in a DNA report?
It means your file has no usable marker at that position, so we have nothing to read. Consumer arrays sample roughly 0.02% of the genome, chosen before anyone looked at you. A gap is an absence of data, not a reassuring result, and reporting it as normal would be the more comfortable lie.
This is the single most common way DNA reports mislead people. A report that lists a gene and shows nothing alarming reads as an all-clear, whether or not the file contained anything to look at.
So coverage is stated per reading. Two pharmacogenomic genes, CYP2D6 and HLA-B, are marked as not readable from array data at all, because that is true of the technology rather than of your particular file.
Which medication markers a consumer file can and cannot show →
Why does DNA Checker not give risk percentages?
Because an absolute number implies a precision the data does not have. A published association is measured in a particular population under particular conditions; restating it as your personal percentage hides how much depends on environment, ancestry and study design. We describe direction and strength of evidence instead.
A figure like "23% likelihood" is memorable, quotable and almost always wrong when applied to one person. The honest version is duller and more useful: what the research found, in whom, how consistently, and how much of the outcome genetics explains at all — which for most common traits is a minority of it.
What happens when studies disagree about a marker?
We leave it out, or we say plainly that the evidence is mixed. A single striking paper is not a finding; replication across populations is. Where a marker is well established for one ancestry and unstudied in others, the report says so, because most published genetics research was done in European cohorts.
That last point is a real limitation of the field rather than a caveat we add for form. Reference panels, array design and the research literature all lean European, so a file from a largely non-European background is often less informative — and any service that does not mention it is glossing over something substantial.
What can a consumer DNA file never tell you?
Anything it has no marker for. Genotyping arrays read pre-chosen positions, so a variant the chip does not cover is simply absent — including most of the rare variants that matter clinically. For genes such as BRCA1 and BRCA2, arrays typically carry only a few founder variants, so a clear reading is not an all-clear.
Anyone with a family history that worries them should be talking to a clinician about proper sequencing, not reading an array file. We say this on the relevant pages rather than in a footer, because it is the situation where a consumer file is most likely to be misread as reassurance.