Raw Genetic Data Interpretation and Third-Party Analysis Services

Start here

Some people download the "raw data" file from a consumer DNA test and upload it to a different website for a new report. This page explains what that file is, how reliable the results are, and what happens to your data. It is education, not advice. If an outside report says you carry a serious variant, do not act on it alone: ask a clinician or genetic counselor about confirmatory testing in a clinical laboratory. See talking with a health professional.

What raw data and third-party analysis are

Third-party genetic interpretation services help people understand raw genetic data obtained from researchers, clinicians or consumer testing companies. Their aims vary: matching users to genetic relatives, selling diet and fitness plans, and giving health risk assessments. [1] A review of five such websites that use SNP-based data found ethical concerns similar to consumer tests, including inadequate informed consent, questionable clinical validity and utility, and lack of medical supervision. [2] Raw files from a genotyping array are a list of base calls at selected positions, not a clinical report. This is one tool in the broader field described in genomics explained.

How accurate is raw data?

Evidence label: mixed. Accuracy depends on how common the variant is.

• Common variants: in 49,908 UK Biobank participants with both array and sequencing data, sensitivity, specificity and predictive values exceeded 99% for 108,574 common variants. [3]
• Very rare variants: for variants below 0.001% frequency, only 16% of 4,757 heterozygous calls were confirmed by sequencing. [3]
• BRCA1 and BRCA2: for rare pathogenic variants in these genes, array sensitivity was 34.6% and positive predictive value 4.2%. [3]
• Consumer data: in 21 people who shared consumer test data online, 20 had at least one false-positive rare pathogenic variant. [3]

A US clinical laboratory reviewing 49 samples sent for confirmation found that 40% of variants reported in raw data were false positives. This was a laboratory case series, not a random sample. Some variants labeled "increased risk" in raw data or by a third-party service were classified as benign by clinical laboratories and are common in population databases. [4]

Two UK genetics centres described patients who used third-party services on raw data, then asked the NHS for interventions based on "pathogenic" variants that were not actually present. [5] That report describes composite cases, so it illustrates a risk rather than measuring how often it happens.

What the services can and cannot tell you

A third-party report matches your variants to published associations. It does not confirm the variant is really present, does not judge how much risk it adds on top of other factors, and is not clinical testing. MedlinePlus notes that consumer test results are generally not meant to diagnose, prevent or treat disease. [6] A variant listed in a research database can also have different meanings in people of different ancestries; see ancestry and population applicability.

Oversight and privacy

MedlinePlus states there is currently limited oversight or regulation of consumer testing companies and of third-party services that analyze genetic information. [7] A 2020 legal analysis said US regulators had not taken an official position on third-party interpretation services, and concluded federal jurisdiction over them is limited but could be appropriate. [1] The five services reviewed in 2017 did not sell or share user data, and three of five did not keep data long term, though that reflects only those sites at that time. [2] Other services may differ, so read each privacy policy. MedlinePlus suggests asking whether you can delete your data, whether it is shared or sold, and what happens if the company is sold. [8]

Uploading genetic files to apps or AI tools adds further privacy questions; see uploading lab or genetic results to apps and AI and genetic privacy and incidental findings. Family members share your DNA, so uploading can affect them too. [8]

How to evaluate a third-party report

• Does it say which variants were actually genotyped, and how confident the call is?
• Is the variant rare? Rare variants on arrays need clinical confirmation. [3]
• Does it claim to diagnose or direct treatment? Consumer reports generally are not meant to. [6]
• Who runs the service, and what does the privacy policy say about sharing and deletion? [8]