Key Takeaways
- A new study reveals genetics significantly influences weight loss outcomes with semaglutide and tirzepatide, highlighting the role of the GLP1R gene variant.
- Approximately one in three patients taking semaglutide may lose less than 5 percent of body weight, while others may lose much more.
- A variant in the GLP1R gene is linked to greater weight loss, but the clinical implications require further research and real-world data.
- Tirzepatide users with a specific GIPR variant face a severely increased risk of nausea, demonstrating the need for personalized treatment strategies.
- Current predictive models combining genetics and clinical factors can explain around 25 percent of weight loss variance, but they’re not definitive.
The numbers are striking, if you stop to look at them. Among people who take semaglutide for weight loss, the average reduction is roughly 10 percent of body weight. Fine. But that average conceals a spread so wide it barely functions as useful information: about one in three patients loses less than 5 percent, or gains weight altogether, while a small fraction shed more than a quarter of their starting weight. Nausea? Some people barely notice it. Others are laid up for weeks. For years, nobody had a good answer for why the same drug, at the same dose, could produce such wildly different journeys in different bodies. A new study, published today in Nature, suggests part of the answer has been sitting in the genome all along.
Researchers at the 23andMe Research Institute conducted what is probably the largest genetic study of GLP-1 medication response ever attempted, drawing on self-reported data from 27,885 people who had taken semaglutide (Ozempic, Wegovy) or tirzepatide (Mounjaro, Zepbound). The scale matters here. Previous pharmacogenetics studies in this area were small, often numbering in the hundreds, and their findings never quite replicated cleanly.
GLP-1 receptor agonists mimic a gut hormone that signals the pancreas to release insulin, slows gastric emptying, and suppresses appetite centrally. Tirzepatide also targets a second incretin receptor, GIP, which may modulate the gastrointestinal side effects that trouble some patients. Individual variation in the receptor proteins these drugs bind to has always seemed like an obvious place to look for explanations. Until now, the datasets weren’t large enough to find anything convincing.
Multiple factors are involved, from drug type and dose to diabetes status and sex, but genetics now appears to play a measurable role. A variant in the GLP1R gene, which encodes the receptor these drugs bind to, is associated with enhanced weight loss, and the combined effect of genetic and clinical factors can predict a spread of outcomes ranging from around 6 to 20 percent of starting weight. The biology isn’t fully explained yet, but the evidence that receptor-level genetic variation matters is now robust enough to replicate in independent datasets.
For people with a particular variant in the GIPR gene, the risk of vomiting on tirzepatide specifically appears substantially elevated. The GIP receptor is thought to buffer the nausea that GLP-1 receptor stimulation tends to cause; if that receptor functions less efficiently due to the variant, the buffer is weaker. The same variant has no detectable effect on semaglutide users, which only targets the GLP-1 receptor, making the drug-specific pattern a meaningful clue about mechanism.
Not with current evidence, but the study points in that direction. The genetic signals identified so far explain only a fraction of the variation in outcomes, and the full clinical utility depends on prospective trials that haven’t been done. What the research does establish is that the question is worth asking, and that the relevant variants are in the very genes these drugs target, which makes the biology unusually plausible.
More reliable than nothing, less reliable than a crystal ball. A combined model using genetics alongside age, sex, drug type, dose, and disease history explained around a quarter of the variation in weight loss in the study population. For comparison, non-genetic factors alone accounted for most of that predictive power; genetics added a modest but genuine increment. The honest answer is that the field is early, the models will improve as datasets grow, and the current tools are better framed as informing expectations than guaranteeing outcomes.
The key finding involves a missense variant in the GLP1R gene, the gene that encodes the receptor semaglutide and tirzepatide both act on. A single nucleotide change at position rs10305420 swaps a proline for a leucine at the seventh amino acid of the protein’s signal peptide, the short sequence that guides the receptor to the cell surface. People carrying the leucine variant lost an estimated 0.76 kg more per copy of the allele, a modest but statistically robust effect (p = 2.9 × 10⁻¹⁰) that was independently replicated in the All of Us cohort using electronic health records.
The variant is most common in people of European and Middle Eastern ancestry (around 40 percent carry at least one copy) and least common in people of African ancestry (about 7 percent). That pattern complicates any clinical application. So does the proposed mechanism: the leucine substitution appears to make the signal peptide more hydrophobic, which may improve the receptor’s stability as it traffics to the cell surface, essentially putting more of the drug’s docking site where the drug can reach it. But “may” and “appears” are doing real work in that sentence. The team found no co-localization with known expression quantitative trait loci in the region, lending weight to the missense interpretation, but the functional experiments haven’t been done yet.
“The approach to weight management is typically one of trial and error,” said Dr. Noura Abul-Husn, Chief Medical Officer at the 23andMe Research Institute. “This can lead people to leap into treatment with a high degree of uncertainty and unrealistic expectations about efficacy and possible side effects.”
The side-effect findings are, in some ways, the more immediately interesting result. Two signals near the GLP1R locus were associated with nausea and vomiting in all GLP-1 users, consistent with the idea that the efficacy variant and the tolerability variants share a causal mechanism (co-localization probability above 88 percent across all three traits). More striking is a completely separate signal: a missense variant in GIPR, the GIP receptor gene that tirzepatide, but not semaglutide, also targets. People carrying the higher-risk version of this variant faced an estimated 1.83-fold increased odds of vomiting on tirzepatide. In the unlucky subset homozygous for risk alleles at both GLP1R and GIPR, that estimated odds ratio for tirzepatide-induced vomiting rose to nearly 15-fold. The variant had no detectable effect on semaglutide users at all, which fits neatly with the hypothesis that GIP receptor activity buffers GLP-1-induced nausea; if your GIP receptor functions poorly, that buffer is largely gone.
Genetics explained a modest share of the overall variance. A combined model incorporating sex, age, drug type, dose, duration, diabetes status, liver disease and hypertension alongside genetic variables accounted for about 25 percent of the variation in weight loss, with most of that coming from the non-genetic factors. For nausea prediction the area under the ROC curve was 65 percent, for vomiting 68 percent; useful, but nowhere near a clinical crystal ball. The authors note that additional data will likely reveal further associations, which is a polite way of saying this is almost certainly the first of several rounds.
There are genuine methodological wrinkles. The study population was 82 percent female and 78 percent of European ancestry. Self-reported weight loss consistently exceeded what electronic health records captured for the same individuals, probably because EHR data misses telehealth prescriptions and provider switches. The directionality of the GLP1R variant effect also contradicts two earlier smaller studies, though both lacked the power to detect the modest effect size, and the All of Us replication in independent EHR data supports the current direction.
23andMe has already translated the findings into a consumer report available through its Total Health service. The tool estimates an individual’s predicted weight loss range and nausea probability based on their genotype alongside demographic and clinical factors, with a clinician involved in delivering the results. Among research participants, the tool generates weight loss estimates ranging from 6 to 20 percent of starting weight, with nausea probability spanning from 5 to 78 percent, depending on the combination of inputs. Whether that information changes prescribing decisions in practice will require prospective clinical data that doesn’t yet exist.
The deeper significance may be less about this particular variant and more about what the study demonstrates is possible. GLP-1 medications are now among the most prescribed drugs in the United States, with roughly one in eight Americans estimated to have used them. The variability in response is enormous, the cost is high, and the side effects drive a significant fraction of people off treatment entirely. Any signal that could help match a patient to the right drug, at the right dose, before the first injection is administered is worth pursuing seriously. The genome may have more to say about that question than the field has so far been able to hear.
Source: Su, Q.J. et al., Nature (2026). https://doi.org/10.1038/s41586-026-10330-z
