A 27,885-Person GWAS Predicts GLP-1 Response And Side Effects From Two Genes. Nature 2026

23andMe Research Institute has published a genome-wide association study of 27,885 GLP-1 users in Nature, April 2026. Variants in two genes predict both weight-loss response and the nausea-vomiting side-effect profile. Senior author Adam Auton, vice president at the 23andMe Research Institute, frames it as “proof of concept that genetics is playing a role in terms of GLP-1 efficacy and side effects.” The route from result to clinic is still open.

The Result

Design:

• 23andMe Research Institute
• Published in Nature, April 2026
• 27,885 GLP-1 receptor agonist users including semaglutide and tirzepatide
• Self-reported weight loss and adverse-event data, GWAS
• Variants in two genes predicted both efficacy and side-effect risk

What the variants predict:

• Whether a user achieves substantial weight loss
• Likelihood of nausea and vomiting, the most common adverse events
• 23andMe will surface the variants on its Total Health platform

Reactions:

• Auton: first concrete evidence that genetics shapes both effect and side effect
• Outside researchers: intriguing, but a meaningful gap remains before clinical adoption
• Early-stage personalized medicine, not yet a prescribing guideline

Why It Matters — Why Some Respond And Others Don’t

GLP-1 receptor agonists rewrote the obesity playbook, but outcomes are not uniform. Roughly 10 to 20 percent of users do not achieve meaningful weight loss, and discontinuation due to nausea and vomiting is common, particularly in the first six weeks. Clinicians have lacked a way to predict either outcome before starting therapy.

This GWAS goes after two clinical questions in one analysis.

The efficacy gap: At identical dosing and dietary support, why does one patient lose 15 percent of body weight at 12 months while another loses under 5 percent? The 23andMe data argue that part of the gap is heritable.

Avoiding side effects: The first four to six weeks of nausea and vomiting are the leading reason for discontinuation. If genetic risk is high, a lower starting dose or slower titration becomes evidence-supported, not just clinical instinct.

Limits — Why It’s Not Yet A Clinical Tool

Self-reported data: All 27,885 weight and side-effect observations are user-reported, not trial-grade measurements. Effect sizes may be over- or under-estimated.

Direct-to-consumer policy: 23andMe will surface variants through Total Health, but most non-U.S. jurisdictions either don’t recognize direct-to-consumer genomics as diagnostic or regulate it separately.

No guideline yet: Neither the Obesity Medicine Association, the European Society of Cardiology, nor the Korean Society for the Study of Obesity recommends pre-prescription genetic testing for GLP-1 therapy. This paper is a first step toward that, not the recommendation itself.

Clinical Read

Pre-prescription genome screening on the horizon: Over the next five to ten years, a simple pre-GLP-1 genomic panel is plausible. Insurance coverage, anti-discrimination policy, and panel cost will gate it.

Other obesity drugs will follow: If GLP-1 response is partly genetic, expect similar GWAS for next-generation agents (GLP-1/GIP/glucagon triple agonists). Drug-to-genome matching is becoming the design pattern.

Polygenic scores will mature: Single-variant prediction will give way to dozens or hundreds of variants combined into polygenic risk scores, with better calibration. The 27,885-person GWAS will feed those models.

What Travels

In most regulated markets, semaglutide and tirzepatide are increasingly available, but pre-prescription genomic screening is not part of the care path. Direct-to-consumer platforms like 23andMe Total Health aren’t universally accessible. Three practical reads.

Use family history: If a first-degree relative responded well or quit GLP-1 therapy due to nausea, share it with your prescriber. Family history is an imperfect but real proxy for the underlying genetic signal.

Negotiate a lower starting dose: If a patient or family history points to nausea risk, ask about a slower titration. Clinical guidelines allow flexibility within the licensed dosing range.

Recalibrate the self-blame frame: “I’m not losing because I lack willpower” doesn’t match the data. Identical dosing produces different outcomes for partly genetic reasons.

What’s Next

23andMe is integrating the variants into Total Health. NIH is reviewing follow-on grant proposals that combine polygenic scores with clinical phenotype data. Critics rightly note that the dataset is heavily European-ancestry; East Asian and African-ancestry cohorts are needed to test cross-population generalization.

The takeaway for now is unfussy. Obesity treatment is drifting away from one-size-fits-all, and the biology that makes the same drug work differently in different people is finally getting visible.