When the Genetic Ruler Is Bent, Everyone Looks Sick

Key Takeaway: A small but striking study revealed that all healthy participants undergoing genome sequencing received false alarms for potentially disease-causing genetic variants. The cause wasn’t errors in the sequencing process but known flaws in the standard reference genome used to interpret their DNA. This finding highlights the urgent need to update the fundamental tools of genomic medicine to prevent millions more from receiving misleading results.

A Perfect Score Nobody Wants

Imagine walking into a clinic for a state-of-the-art preventive health screening—whole-genome sequencing, the kind marketed as a window into your future disease risks. Now imagine being told your results flag potentially serious genetic variants. You’d be worried. You’d lose sleep. Your doctor might order additional tests, refer you to a genetic counselor, or even begin surveillance for a disease you will never develop. Now imagine that every single person who took the same test received the same false alarm. That’s exactly what happened in a recent study, and its implications should concern anyone who has had, or is considering, genomic testing.

What the Researchers Did

In an observational case series study, researchers sequenced the genomes of 20 healthy adults using standard clinical-grade methods. The goal was simple: to examine how automated bioinformatics pipelines—the software that interprets raw DNA data—identify genetic variants in the context of a preventive screen. The sequencing data was aligned against GRCh38, the current standard human reference genome maintained by the Genome Reference Consortium, which has served as the global benchmark for interpreting human DNA since its publication in 2013^[2].

What They Found

The results were staggering. One hundred percent of participants—20 out of 20—received false-positive flags for “high-impact” genetic variants^[1]. These weren’t randomly distributed errors. The false alarms clustered in three specific regions of the genome: the genes SLC37A4, CIMIP2A, and GPR33. When the researchers cross-referenced the flagged variants against gnomAD, one of the world’s largest databases of human genetic variation containing data from hundreds of thousands of individuals, the picture became clear^[3]. The alleles flagged as dangerous were, in fact, the most common versions of these genes in the global population—alleles that are entirely benign and carried by the vast majority of healthy people.

The Mechanism: When the Ruler Itself Is Flawed

To understand how this happens, you have to understand what a reference genome actually is. When your DNA is sequenced, the machine doesn’t read it like a book from cover to cover. Instead, it chops your genome into millions of tiny fragments, and a computer then reassembles these pieces by comparing them to a template—the reference genome. Any spot where your DNA differs from the template is marked as a “variant.” Some variants are harmless. Others are classified as potentially pathogenic, meaning they could contribute to disease.

The problem is this: the reference genome isn’t a perfect representation of a typical human. It was originally built from a small number of donors and, in certain locations, contains what geneticists call “reference minor alleles” (RMAs)^[4]. These are versions of a gene that are actually rare in the global population but happen to be the ones included in the reference. Think of it this way: if your measuring stick is slightly warped, every wall you measure will look crooked to you, even if the walls are perfectly straight.

At the three loci identified in this study, the reference genome carries rare alleles. So, when the DNA of a healthy person carrying the common, normal version is compared against this flawed template, the software interprets the common allele as a deviation. The automated calling tools then label it as “high-impact,” triggering a clinical alarm that has no biological basis. The sequencing was correct. The interpretation infrastructure was not.

This is not a theoretical concern. One of the three genes involved, SLC37A4, is associated with glycogen storage disease type Ib, a serious metabolic disorder^[5]. A false-positive flag at this locus could trigger a cascade of unnecessary anxiety, specialist consultations, and invasive testing for a variant that most people carry without consequence.

Why This Matters Now

Genomic screening is no longer confined to research hospitals. Direct-to-consumer testing has exploded, and preventive genome sequencing is rapidly entering mainstream clinical practice, particularly in longevity and geroscience medicine. As these systems scale to millions of people, systematic errors embedded in the reference genome become not just individual patient issues, but population-level problems. Every false alarm consumes clinical resources, generates patient anxiety, and erodes trust in a technology with real potential.

The authors argue the solution lies in transitioning from the current linear reference genome to graph-based pangenome references. Unlike a single linear template, a pangenome reference incorporates genetic variation from diverse populations, representing multiple common alleles at each position rather than a single arbitrary choice^[6]. The Human Pangenome Reference Consortium has already released draft pangenome references built from dozens of diverse individuals, and initial analyses show these tools significantly reduce reference bias^[7].

Limitations to Consider

This was a small case series of 20 participants, and the false positives were concentrated at just three genomic loci. The study does not claim that every variant call in clinical genomics is unreliable—the vast majority of the genome is well-represented by GRCh38. But the fact that these three loci produced false positives 100% of the time in healthy individuals highlights a systematic class of error that likely extends to other RMA regions scattered throughout the genome. Larger studies are needed to map all such vulnerable positions.

Conclusion: What This Means for You

If you’ve had genome sequencing and received alerts about potentially harmful variants, this study is a reminder that context is everything. A flagged variant is not a diagnosis—it’s a computational call that is entirely dependent on the quality of the tools used to generate it. For clinicians ordering genomic screens, the message is equally clear: automated systems require a layer of manual curation, especially at known RMA loci, before results reach patients. And for the field of preventive genomics as a whole, the clock is ticking. Every month that passes without adopting pangenome references is another month of preventable false alarms reaching real people making real decisions about their health.

Scientific Sources

1. Nagy GR, et al. Beyond the linear genome: how reference bias threatens preventive medicine and geroscience. GeroScience. 2026. PubMed: https://pubmed.ncbi.nlm.nih.gov/42501271/
2. Schneider VA, et al. Evaluation of GRCh38 and de novo haploid genome assemblies demonstrates the enduring quality of the reference assembly. Genome Res. 2017. DOI: 10.1101/gr.213611.116
3. Karczewski KJ, et al. The mutational constraint spectrum quantified from variation in 141,456 humans. Nature. 2020. DOI: 10.1038/s41586-020-2308-7
4. Ballouz S, et al. Is it time to change the reference genome? Genome Biol. 2019. DOI: 10.1186/s13059-019-1774-4
5. Chou JY. The molecular basis of type 1 glycogen storage diseases. Curr Mol Med. 2001. DOI: 10.2174/1566524013364112
6. Eizenga JM, et al. Pangenome graphs. Annu Rev Genomics Hum Genet. 2020. DOI: 10.1146/annurev-genom-120219-080406
7. Liao WW, et al. A draft human pangenome reference. Nature. 2023. DOI: 10.1038/s41586-023-05896-x