Third-party raw-data reports throw false alarms for three main reasons: the underlying test can misread rare variants, the software matches variants to databases without clinical judgment, and many "risk" labels describe weak or uncertain evidence. A concerning result is a reason to confirm, not to act.
Two peer-reviewed studies show the scale of the problem, and they point to the same rule: confirm in a clinical laboratory before anything changes in your care.
What the studies found
Clinical lab retesting. A 2018 study in Genetics in Medicine from Ambry Genetics looked at 49 patients whose consumer or third-party results were sent for clinical confirmation. Of 43 variants reported, 17 (40%) were not found on retesting, and eight reported as "increased risk" were classed as benign, common or of uncertain significance by the clinical lab. The sample was small and only included people who sought confirmation, so it is not a population error rate. The authors' conclusion was that all consumer tests need confirmation by a clinical laboratory.
Population data. A 2021 evaluation in The BMJ compared genotyping chips with sequencing in almost 50,000 UK Biobank participants. For BRCA1 and BRCA2 variants called pathogenic by the chips, the overall positive predictive value was 4.2%: of 889 participants flagged, only 37 were confirmed by sequencing. The authors advised that chip results for very rare variants should not guide health decisions without validation.
Why rare variants are the weak spot
Genotyping arrays read a fixed list of positions, and they are built mainly for common variants (see whole genome sequencing versus genotyping). When an array reports a variant that very few people carry, even a small error rate means most of the "hits" are mistakes, because genuine carriers are so scarce. That is why the BMJ team found chip accuracy varied widely by chip type for the same genes.
Why third-party software adds more noise
- It matches, it does not interpret. Many tools compare your variants with public databases and print everything that appears. Some reports are deliberately unfiltered, listing trivial and important findings together. Our guide to raw-data tools describes how different services approach this.
- Evidence varies. A database entry can rest on one small study or conflicting submissions. A label of "associated with" or "increased risk" is not the same as a clinical classification of pathogenic.
- Classifications change. Variants are reclassified as evidence accumulates, often downward. See variants of uncertain significance explained.
- Population matters. A variant that is rare in one ancestry can be common in another, and a report built on a limited reference can mislabel it.
- Annotations are snapshots. A report reflects the database version the tool used on the day it ran, so the same file can produce a different list a year later.
- Quality flags get ignored. A VCF's FILTER and depth fields show how confident the caller was (see opening your VCF). Reports may not show a weak call as weak.
The pattern across both studies is consistent. The alarming results tend to be the rarest ones, they are the ones most likely to be wrong, and they are the ones people are most tempted to act on. A confirmed finding from a clinical laboratory looks different: it comes with a formal classification, a stated method and a clinician who can explain what follows.
A five-step triage for a scary result
- Read the exact label. "Pathogenic" or "likely pathogenic" is very different from "risk factor" or "uncertain significance."
- Note the method. A chip result for a rare variant carries the most doubt. A whole genome call with high depth is better but still not a clinical diagnosis.
- Check the quality. In a VCF, look at the FILTER value, read depth and genotype quality for that line.
- Look up the variant's evidence. Public resources such as ClinVar show classifications and how many labs agree. A high frequency in the general population argues against a rare severe condition.
- Get it confirmed. Ask a clinician for confirmatory testing in a clinical lab. Our guides on bringing results to your doctor and seeing a genetic counselor explain how.
Do not stop or start medication, or plan surgery, on the strength of a raw-data report. And do not alarm relatives until the finding is confirmed. See how to tell your family once it is.
Sequencing reads the letters, then you confirm what matters
Whole genome sequencing from Dante Labs gives you raw data with full coverage of your genome. Treat any serious finding as a lead and confirm it with your doctor.
Compare Dante Labs WGS → Code GENOME — 10% offAffiliate link. Consumer sequencing is not a clinical diagnostic. Confirm medically important findings through your doctor.
Does whole genome sequencing fix this?
It helps, because sequencing reads the actual letters instead of inferring them, and in the BMJ study sequencing was the reference standard. But better data does not remove interpretation uncertainty, and a consumer report is still not a clinical diagnosis. Our DTC versus clinical-grade testing guide and the hereditary cancer panel versus WGS comparison show when a clinical test is the better tool.
Bottom line
A raw-data alarm is a lead, not a verdict. Most of the time, careful triage and one clinical confirmation settles it. Stay calm, gather the facts and let a professional confirm before acting.
Common questions
How often are raw-data results wrong?
It depends on the variant and method. In one 2018 study of people who sought clinical confirmation, 17 of 43 reported variants (40%) were not confirmed. In a 2021 population study, genotyping chips had a positive predictive value of only 4.2% for rare BRCA variants.
Should I worry if a tool flags a BRCA or cancer variant?
Take it seriously enough to confirm it, but do not act on it alone. Ask a clinician about clinical-lab testing, and consider a genetic counselor.
What is the difference between pathogenic and risk factor?
Pathogenic means strong evidence that a variant causes disease. A risk factor usually means a statistical association that raises the chance modestly and does not guarantee a condition.
Is whole genome data more reliable than a chip?
It reads the actual sequence, so it tends to be more reliable for rare variants. It still needs quality filtering, careful interpretation and confirmation for anything medically significant.
Can I trust a free raw-data analysis tool?
Treat it as exploratory. Free and low-cost tools can be useful for curiosity, but they are not clinical tests, and their output needs verification before it affects medical decisions.
Sources and verification
- Tandy-Connor et al., Genetics in Medicine (2018): false-positive results released by direct-to-consumer genetic tests highlight the importance of clinical confirmation testing
- FORCE (Facing Our Risk of Cancer Empowered): summary of the Tandy-Connor study (49 cases, 43 variants, 17 false positives)
- Weedon et al., The BMJ (2021): use of SNP chips to detect rare pathogenic variants
Study figures were read from the sources above in October 2026. They describe specific samples and chips and should not be read as a general error rate. The triage steps are general good practice, not medical advice for your case.
Medical disclaimer. GenomeTesting.org publishes educational content about genetics and consumer genomic testing. Nothing here is medical advice, a diagnosis, or a substitute for care from a qualified clinician or genetic counsellor. Decisions about testing, treatment, or interpreting your own results should be made with a healthcare professional who can see your full history.
Affiliate disclosure. Some links on this page are affiliate links. If you buy through them we may earn a commission at no extra cost to you. This never changes which products we recommend or what we say about them.