What a consumer DNA chip can and cannot tell you
Tests such as 23andMe, AncestryDNA, MyHeritage and FamilyTreeDNA read your DNA with a genotyping chip. A chip is good at common variants and weak at rare ones. Knowing the difference tells you which results to take seriously, which need a clinical test, and why an empty section does not mean you are in the clear.
- Positions read
- between 574,516 and 992,615 chosen positions, depending on the chip version — not your whole genome
- Rare variants
- often wrong on a chip; a rare disease-causing variant in raw data needs a clinical test before anything is done about it
- BRCA1 and BRCA2
- not assessed from a chip by greatgenes; a negative chip result does not rule out a hereditary cancer variant
- Polygenic scores
- a comparison with other people, not a diagnosis and not your personal risk
What a chip reads well
In short: a chip measures a fixed list of positions where people commonly differ. For common variants chip readings of them are mostly confirmed by sequencing (Weedon et al., BMJ 2021).
A genotyping chip does not read your DNA letter by letter. It checks a set list of positions chosen by the chip maker, mostly common single-letter differences (SNPs) useful for ancestry and relative matching. Consumer chips read between 574,516 and 992,615 such positions, depending on the chip version.
That is enough for well-studied common variants: many medicine-related variants, common traits, and the many small effects that polygenic scores add up. greatgenes recognises the chip version from the file and uses only positions that chip actually measured.
Rare variants: the main source of false alarms
In short: the rarer a variant, the more likely a chip reading of it is wrong. Many rare disease-causing variants reported from raw chip data are not confirmed by a clinical test, and for the rarest variants most are not.
- In UK Biobank, chip results were compared with sequencing of the same people. For variants with a frequency below 0.001 %, only 16 % of 4,757 chip calls were confirmed; for pathogenic BRCA1 and BRCA2 variants the positive predictive value was 4.2 % (Weedon et al., BMJ 2021).
- A clinical laboratory did not confirm 40 % of the variants from direct-to-consumer raw data that were sent to it for testing (Tandy-Connor et al., Genet Med 2018).
greatgenes takes this into account. If the chip reports a pathogenic variant that is rare (below 0.1 % in the population) or has no known frequency, it is not among the main findings in the Summary, its reliability is shown as “low — needs confirmation”, and the report says that a clinical confirmation test is needed before any conclusion; for two copies it also says that this is a likely chip error. Details are in How greatgenes works: finding priority and reliability.
BRCA1, BRCA2 and other cancer genes
In short: hereditary cancer genes can carry any of a very large number of rare variants, and a chip tests only a handful of them, unreliably. A “nothing found” from a chip is not a negative genetic test.
- When the first direct-to-consumer BRCA report was authorised in the US, it covered three pathogenic variants, while there are over a thousand known pathogenic BRCA1 and BRCA2 variants (Roberts et al., JAMIA Open 2019).
- greatgenes does not assess BRCA1 and BRCA2 from a chip, and the report says so.
- The “Checked, no signs found” section covers the pathogenic and likely pathogenic ClinVar variants at positions your chip reads, including those in the 81 genes of the ACMG SF v3.3 list. It always says “no signs in your data”, never “you do not have the disease”.
If cancer runs in your family, ask a doctor or a genetic counsellor about clinical testing, whatever a chip shows.
What a chip does not read at all
- Positions that are not on the chip. Most of the genome is not measured. If a variant is not in your file, the report cannot say anything about it.
- Most insertions and deletions. Chips encode them unreliably, and greatgenes does not use them. Some medicine genes have common variants of this kind (in CYP3A5 and CYP2D6, for example): a gene copy that carries one looks normal on a chip. The report allows for such variants by their frequency in populations and then says “likely …” or “not determined” instead of a firm result.
- The number of gene copies. This matters, for example, for the medicine gene CYP2D6: a chip cannot see whether you have extra or missing copies. On some chip versions the file also lacks the position of the most common non-working CYP2D6 variant in people of European ancestry, and then the report does not assess CYP2D6 at all: 23andMe v2; 23andMe v5; AncestryDNA v1 / FamilyTreeDNA (OmniExpress); AncestryDNA, array c5.0.0; AncestryDNA, array c5.1.0; Illumina GSA v5.1.0 (FamilyTreeDNA, 23andMe).
- Larger structural changes and repeat expansions.
- Which copy of a chromosome a variant sits on. A chip gives two letters per position but does not say which came from which parent.
For medicines this means that a “normal” result from a chip is not a clinical test. For example, the US Food and Drug Administration advises DPYD testing before treatment with capecitabine or fluorouracil (FDA safety labeling update), and a chip reads only some of the DPYD variants that matter. How the report handles medicine genes: How greatgenes works.
No imputation
In short: greatgenes uses only what your chip measured. It does not guess missing positions.
Some services “fill in” unmeasured positions by estimating them from reference panels (imputation). Imputed genotypes are estimates, and they are least accurate for exactly the rare variants that matter most medically. greatgenes does not impute. As a result some modern polygenic scores, built for imputed genomes, are not used, and every genotype in your report is one your chip actually read.
Polygenic scores are comparisons, not diagnoses
In short: a polygenic score places you relative to other people. Most people with a high score never develop the condition, and many people with a low score do.
- A score adds up many small effects of common variants. It is compared with a reference population, so the result is a position (for example “above average”), not a yes or no.
- greatgenes uses 15 published scores and assesses a score only if your chip covers at least 50 % of it.
- Lifestyle, family history and age also matter, and the score does not include them.
- Scores are less accurate for people whose ancestry differs from the people the score was built on — mostly people of European ancestry.
More on how scores work: What is a polygenic score.
What to do with a worrying result
- Read the reliability word next to the finding and whether the report flags it as a possible chip error.
- Do not make medical decisions on the basis of raw data. A result is not a diagnosis.
- Discuss the finding with a doctor or a clinical geneticist and ask whether a clinical test is needed to confirm it.
- Never start, stop or change a medicine on your own; show a medicine-related result to your doctor or pharmacist.
Analyse your file — in your browser, without uploading it.
References
- Weedon MN et al. Use of SNP chips to detect rare pathogenic variants: retrospective, population based diagnostic evaluation. BMJ 2021. doi:10.1136/bmj.n214
- Tandy-Connor S et al. False-positive results released by direct-to-consumer genetic tests highlight the importance of clinical confirmation testing for appropriate patient care. Genet Med 2018. PMID 29565420
- Roberts MC, Allen CG, Andersen BL. The FDA authorization of direct-to-consumer genetic testing for three BRCA1/2 pathogenic variants: a twitter analysis of the public’s response. JAMIA Open 2019. doi:10.1093/jamiaopen/ooz037
- ClinVar: ACMG recommendations for reporting of secondary findings (NCBI).
- Martin AR et al. Clinical use of current polygenic risk scores may exacerbate health disparities. Nat Genet 2019. doi:10.1038/s41588-019-0379-x