23andMe Ehlers-Danlos Insights: Reading Your Genetic Data

Direct-to-consumer genetic tests like 23andMe were not designed to diagnose Ehlers-Danlos syndrome, and the raw data they produce covers only a tiny fraction of the genetic territory that EDS spans. Twelve of the thirteen recognized EDS subtypes are caused by variants in any of 21 different genes, and the most common subtype, hypermobile EDS, has no confirmed genetic cause at all. That mismatch between what the test measures and what the condition requires makes 23andMe data a poor lens for understanding your EDS risk, though the raw files can occasionally surface something worth investigating with a clinical geneticist.

What EDS Looks Like Genetically

Ehlers-Danlos syndrome is not a single disease. The 2017 international classification recognizes thirteen subtypes, twelve of which are monogenic conditions tied to pathogenic variants across 21 confirmed genes.

¹ These genes mostly encode collagens or the enzymes that process them. Classical EDS, for example, typically results from mutations in the COL5A1 or COL5A2 genes, which provide instructions for type V collagen.² Vascular EDS, the most dangerous subtype, involves COL3A1, which encodes type III collagen found in blood vessel walls and hollow organs. Other subtypes involve genes responsible for collagen folding, crosslinking, or the extracellular matrix scaffolding that holds connective tissue together.

Importantly, the 2017 classification emphasizes that for every subtype except hypermobile EDS, a definitive diagnosis depends on identifying the causative genetic variant, not just clinical features.³ This means genetic testing is not optional for most subtypes; it is part of what confirms the diagnosis. The testing required, however, is clinical-grade gene sequencing or panel testing, not a consumer SNP chip.

Why 23andMe Raw Data Misses Most EDS Variants

The core issue is the technology. 23andMe and similar services use genotyping arrays, sometimes called SNP chips, which read a pre-selected set of common genetic positions across your genome. Think of it like checking specific addresses on a street rather than reading every house number. The chip captures around 600,000 to 700,000 positions out of more than three billion base pairs in the human genome. The positions chosen are biased toward common variants, ones that appear frequently in the general population and are useful for ancestry analysis, trait prediction, or well-studied disease risks.

EDS-causing mutations, by contrast, are overwhelmingly rare. Many are private to a single family. A collagen gene mutation that causes classical or vascular EDS might appear in only a handful of people worldwide. These variants are almost never included on a consumer genotyping chip because the chip was not designed to find them. Clinical EDS genetic testing uses targeted gene panels or exome sequencing, which reads the protein-coding portions of DNA base by base. That approach can detect novel point mutations, small insertions, deletions, and even synonymous mutations that alter gene splicing. A SNP chip cannot do any of this.

Some people download their raw 23andMe data file and upload it to third-party interpretation services like Promethease, Livewello, or similar tools. These services search the raw genotype data for positions that have been linked to diseases in databases. Occasionally, the raw file will flag a position in a collagen gene or another EDS-related gene. The question then becomes whether that flag is real and whether it means anything clinically.

The False Positive Problem

When researchers have taken variants flagged as pathogenic in direct-to-consumer raw data and confirmed them with clinical-grade sequencing, the results are sobering. One study found that about 40% of variants reported as disease-causing in DTC raw data turned out to be false positives when checked with validated methods.⁴ The problem gets worse with rarer variants. A large retrospective analysis using UK Biobank and Personal Genome Project data found that for very rare variants, those appearing in fewer than one in 100,000 people, only about 16% of heterozygous genotypes from SNP chips were confirmed by sequencing. Nearly all individuals in the study had at least one falsely flagged rare pathogenic variant.⁵

EDS-causing variants fall squarely into the “very rare” category. If your raw data flags a variant in COL3A1 or COL5A1, the statistical likelihood that it is a genotyping error is high. That does not mean you should ignore it entirely, but it does mean you should not treat it as a diagnosis or panic before it has been confirmed by a clinical laboratory using sequencing-based methods.

Hypermobile EDS Cannot Be Found in Any Genetic Test

The most common form of EDS that people search for, hypermobile EDS, presents a unique frustration: there is no known gene for it. Despite multiple international research efforts, the molecular genetic cause remains unidentified, and inter-family genetic differences make the search even harder.⁶⁷ This means no genetic test, clinical or consumer, can confirm or rule out hEDS. Diagnosis relies entirely on the 2017 clinical criteria, which evaluate joint hypermobility, systemic features of connective tissue fragility, family history, and the exclusion of other conditions.⁸

If you are hoping 23andMe data will tell you whether you have hEDS, the honest answer is that it cannot. No current genetic test can. Patients with symptomatic joint hypermobility who do not fully meet the 2017 criteria are often labeled with hypermobility spectrum disorder, a related but separate classification that is not universally recognized across all healthcare systems.⁸ The distinction between hEDS and HSD matters for research and insurance coding, but the day-to-day management and symptoms overlap considerably.

Research into what drives hEDS at the molecular level continues. One avenue involves epigenetics. A study examining skin cells from hEDS patients found altered expression of multiple microRNAs, small molecules that regulate gene activity without changing the DNA sequence itself. The researchers noted that these epigenetic shifts correlated with changes in gene expression in hEDS cells, suggesting that the condition may involve regulatory disruption rather than a single broken gene.⁹ If that turns out to be the case, it would help explain why standard gene sequencing has come up empty: the problem may not be in the DNA code but in how it is read.

When Raw Data Does Flag Something Real

Despite the noise, it is not impossible for 23andMe raw data to flag a genuine variant in an EDS-related gene. The chip does include some positions in collagen genes, and a small number of known pathogenic variants happen to fall on probed positions. If a third-party tool flags a variant classified as pathogenic in ClinVar or another curated database, the appropriate response is not to self-diagnose and not to dismiss it, but to bring the finding to a genetics professional.

Variant interpretation follows a five-tier classification system developed by professional genetics organizations: pathogenic, likely pathogenic, uncertain significance, likely benign, and benign.¹⁰ Many variants flagged by third-party tools land in the “uncertain significance” category, meaning there is not enough evidence to say whether the variant causes disease. A clinical geneticist or genetic counselor can evaluate whether confirmation testing is warranted, whether the variant has been reclassified since it was first entered in a database, and whether your clinical picture matches what the variant would predict.

The practical challenge is getting people to take that step. One study found that only about 4% of direct-to-consumer testing customers made or planned an appointment with a genetic counselor after receiving their results, though roughly 38% said they would have used in-person counseling if it had been available.¹¹ The gap between wanting guidance and actually accessing it leaves many people sitting with ambiguous raw-data findings and no expert to help them interpret what they are seeing.

The Emotional Weight of Unverified Findings

Discovering a flagged variant in your raw data, especially one linked to a serious condition like vascular EDS, can be deeply unsettling. Qualitative research on patients who found pathogenic or likely pathogenic variants through raw data interpretation found that nearly all described emotional distress, often coupled with intense information-seeking as a coping mechanism.¹² People read medical literature, joined online communities, and sometimes made significant health decisions before the variant had been confirmed or evaluated by a professional.

On the provider side, genetics healthcare professionals report that counseling patients who arrive with consumer genomic results is frequently complicated. In one survey, 94% of genetics providers had encountered at least one challenge in these consultations, including adverse emotional reactions in patients, incorrect variant interpretation by third-party tools, and results that could not be confirmed on clinical-grade testing.¹³ Unconfirmed results were especially common in oncology settings, but the same dynamic applies to connective tissue disorders: a patient walks in convinced they carry a pathogenic variant, and the clinical lab either cannot confirm the finding or finds the variant is benign.

None of this means people are wrong to look at their raw data or to be proactive about their health. It does mean the path between “my third-party report flagged something” and “I have a confirmed genetic diagnosis” is longer and more uncertain than most people expect. Treat a raw-data flag as a question, not an answer.

What Clinical EDS Genetic Testing Actually Involves

If a clinician suspects a monogenic form of EDS, the standard approach is a targeted gene panel that sequences the relevant genes base by base. For classical EDS, that means full sequencing of COL5A1 and COL5A2, plus deletion/duplication analysis.² For vascular EDS, COL3A1 is the primary target. Broader connective tissue disorder panels may include dozens of genes to capture rarer subtypes and conditions that overlap with EDS clinically.

Even clinical-grade testing does not always provide a clean answer. A study using whole exome sequencing on 174 EDS patients who had already undergone expert clinical evaluation, laboratory testing, and targeted gene sequencing but remained without a genetic diagnosis underscores this point.¹⁴ Some patients with clear clinical EDS simply do not have a detectable mutation in a known gene. The researchers examined loci from linkage studies, transcriptome analyses, and genome-wide association data in an effort to uncover new candidate genes. This kind of deep investigation is what the research frontier looks like for EDS genetics, and it is far beyond anything a consumer genotyping chip can approach.

Some rarer EDS subtypes also illustrate just how unusual the causative genetics can be. A recessive form of EDS was identified as being caused by deficiency of tenascin-X, a protein that is not a collagen at all but rather plays a role in maintaining the structural integrity of the collagen matrix. Patients in the original study had homozygous deletions or truncating mutations in the tenascin-X gene, confirming that factors well outside the collagen family itself can cause EDS.¹⁵ These findings reinforce how genetically heterogeneous EDS is. A simple scan of collagen genes would miss this subtype entirely.

Vascular EDS and Why Early Detection Matters

Among the EDS subtypes, vascular EDS is where genetic confirmation carries the highest stakes. People with vascular EDS face a risk of arterial rupture, organ perforation, and other life-threatening complications. Knowing the diagnosis changes medical management significantly. Retrospective data from the UK’s national diagnostic service in Sheffield found that patients on long-term blood pressure medications, specifically angiotensin II receptor blockers or beta-blockers, had fewer vascular events than patients who received only lifestyle and emergency care advice. Five-year survival was roughly 97% for those on medication compared with about 43% for those in a control group receiving no medication.¹⁶

Those numbers are striking enough that a missed or delayed diagnosis of vascular EDS has real consequences. If 23andMe raw data happens to flag a COL3A1 variant, even knowing the high false-positive rate, confirmation testing is urgent rather than optional. A true positive finding in COL3A1 changes your surveillance schedule, your surgical risk profile, and your medication regimen. It is one of the rare cases where a consumer genomics flag, if followed up properly, could be genuinely life-saving.

Social Media, Self-Diagnosis, and the Path to Clinical Evaluation

For many people, the road to an EDS evaluation does not start with genetic testing at all. It starts with a social media post. Research on hEDS patients and social media found that some first learned about the condition through platforms like TikTok, describing the experience as a “eureka” moment when a video’s description matched their symptoms. These individuals then brought the possible diagnosis to a primary care physician and received clinical evaluation.¹⁷ In one case, a participant posted symptoms to a social media community and had hEDS suggested as a possible diagnosis within 72 hours.

This dynamic is relevant to the 23andMe question because it reflects how people actually arrive at EDS as a possibility. Some come from clinical symptoms, some from social media recognition, and some from browsing their raw genetic data. All three paths can be legitimate starting points, but none of them is a diagnosis. The risk with raw data, as with social media identification, is confirmation bias. Once you have a possible label, it is easy to interpret every symptom through that lens and harder to consider alternative explanations. A genetics provider or rheumatologist experienced with connective tissue disorders can evaluate the full picture in a way that a raw data report or a TikTok video cannot.

For the monogenic subtypes, a clinical genetics referral can settle the question definitively through proper sequencing. For hEDS, where no gene test exists, the evaluation is clinical and systematic, using the 2017 criteria and ruling out other connective tissue conditions. Either way, the consumer genetics file is a starting point at best, one with a high noise-to-signal ratio that requires professional interpretation before it means anything actionable.

What Might Change in the Future

Consumer genomics technology is evolving. Some newer services are beginning to offer limited sequencing rather than pure genotyping, and research-grade whole genome sequencing has dropped in cost enough that consumer availability is plausible within years rather than decades. If consumer platforms eventually provide sequencing-level data, the false-positive problem would shrink substantially, and the ability to detect rare EDS variants would improve.

The bigger unknown is hEDS. If the gene or genes responsible are eventually identified, genetic testing for the most common EDS subtype could become possible for the first time. The epigenetic findings already discussed suggest the answer might be more complex than a single gene, potentially involving regulatory mechanisms that standard gene panels would not detect. Whole exome and whole genome approaches being applied in research settings represent the current frontier.¹⁴ Until those efforts produce actionable discoveries, the genetic landscape of hEDS remains a blank spot on the map, for consumers and clinicians alike.