On any given morning, skyrocketing numbers of people reach for a small injection pen (and soon a pill) that, just a few years ago, was barely available outside of diabetes clinics. Drugs like semaglutide and tirzepatide have become cultural phenomena, reshaping not only medicine but also public discourse and the advertising industry around weight, metabolism, and obesity. Today, it is impossible to open a magazine, turn on the TV or radio, or walk down the grocery aisle without encountering some form of advertisement for these GLP-1 receptor agonists (GLP-1RAs).
Almost any individual in the United States can obtain a subscription to a GLP-1RA without having to visit a doctor’s office. Just visit Hims/Hers, Ro, or Noom and answer a few questions about weight, height, goals, and concerns to get a prescription. (One such site claims it is taking weight and height data and “combining with clinical data,” whatever that means, before presenting a plan and steps for ordering a prescription.)
But there are some major problems, one being that these drugs don’t work uniformly. Some patients respond to GLP-1RAs almost immediately, reporting diminished cravings within days. Others see little change. Side effects, too, can vary dramatically, from mild discomfort to debilitating nausea and vomiting. The spread of outcomes is wide and not fully understood.
Before blindly beginning to take a drug that, on the one hand, has seemingly miraculous effects and, on the other hand, might cause serious side effects like pancreatitis, gallbladder disease, and kidney failure, wouldn’t prescribers and prescription seekers want to know this information?
A study published in Nature Medicine by the 23andMe Research Institute—the new nonprofit entity founded by the company’s co-founder, Anne Wojcicki, for $305 million to replace the bankrupt biotechnology company—suggests the answer may be found, at least in part, in something far more fundamental than diet or willpower: our genes.
In speaking with Inside Precision Medicine for the first time since the company filed for bankruptcy and was resold to the nonprofit public benefit corporation, Adam Auton, PhD, vice president of Human Genetics at the 23andMe Research Institute, said, “The ‘GLP-1s’ have completely transformed weight loss management. A huge fraction of the population is benefiting. It’s a very natural question: Are people’s experiences on GLP-1s modulated by genetics?”
The short answer is, yes. Auton and 23andMe Research Institute scientists have provided genetic evidence that variation in drug target genes contributes to variability in response among individuals, laying the groundwork for consumer-based precision medicine approaches to obesity treatment and beyond.
Crowd-sourcing GLP-1 genetics
To better understand why responses to GLP-1 receptor agonists vary so widely, the 23andMe Research Institute team leveraged its uniquely large and engaged research cohort. Over the past decade, the company has assembled genetic data from more than 15 million participants who consented to research, enabling analyses that would be difficult in traditional clinical trials. Immediately following the company’s filing for bankruptcy in March 2025, 23andMe reported that over 1.9 million users requested for their data to be deleted. Auton told Inside Precision Medicine that the current number of consented customers is around 11 million.
Building on this resource, Auton and colleagues launched a targeted survey asking participants detailed questions about their GLP-1 drug use, including medication type, duration, dosage, weight loss, and side effects. More than 27,885 customers responded, providing a rich, real-world dataset. “That’s the power of having a large, engaged cohort,” said Auton. “You can ask a question and very rapidly get meaningful data back.”
Using these data, Auton and colleagues conducted a genome-wide association study (GWAS), scanning millions of genetic variants to identify those associated with treatment outcomes. “You’re starting with the entire genome,” Auton explained. “You’re testing every variant for correlation with the trait of interest. And when you see a signal, it tends to be overwhelming.”
The team focused on two primary traits: weight loss and the presence of side effects. The strongest association emerged in GLP1R, the gene encoding the GLP-1 receptor—the direct target of these drugs. A missense variant, rs10305420, was linked to significantly greater weight loss, with each copy associated with an additional 0.76 kilograms lost.
“It made very clear biological sense,” Auton said. “This is the receptor that the drug is acting on.” The missense variant may affect how much receptor is expressed on the cell surface, meaning individuals with more receptors could experience a stronger response to the same dose.
A second key finding involved a substitution in GIPR (rs1800437; p.Glu354Gln), which encodes the receptor for glucose-dependent insulinotropic polypeptide and is targeted by dual agonists such as tirzepatide. Unlike the GLP1R result, this association was not related to weight loss but to drug tolerability. Carriers of the variant were more likely to report nausea and vomiting—but only when taking medications that act on the GIP receptor. No such effect was observed among users of semaglutide, which does not target GIPR.
“It was very, very clean,” Auton said. “We saw this effect specifically in people taking the medications that actually target that receptor.”
Together, these findings underscore a central principle of pharmacogenetics: genetic variation can shape not only whether a drug works, but also how it is experienced, often in highly drug-specific ways.
Who is represented
One of the study’s more unconventional aspects is its reliance on self-reported data, a method sometimes viewed with skepticism in clinical research given the limits of memory and potential inaccuracies in reporting weight loss or medication use. Anticipating this concern, scientists at the 23andMe Research Institute validated their findings using a subset of participants who also shared electronic health records (EHRs), enabling direct comparison between self-reported and clinically recorded data.
The results were reassuring: survey-reported weight loss closely tracked with medical records, and medication histories aligned well across both sources. Although participants tended to slightly overestimate weight loss, they also reported longer treatment durations, effects that largely offset each other. Importantly, the genetic associations remained robust under independent scrutiny, with replication in the All of Us Research Program, a large, federally funded dataset based on clinical records rather than self-report.
While weight loss is the headline feature of GLP-1RAs, side effects often determine whether patients persist with treatment. Nausea, vomiting, and gastrointestinal discomfort are among the most common reasons for discontinuation, yet they are frequently underreported in traditional clinical datasets. EHRs may document when a medication is stopped but rarely capture why. Self-reported data addresses this gap by directly capturing patient experience.
“We were able to ask people directly about their experiences,” Auton said. “That’s something that’s often missing from clinical datasets.” By linking these experiences to genetic variation, the study enables a more refined understanding of drug tolerability, moving beyond population averages to individualized risk profiles.
As with many large-scale genetic studies, statistical power was greatest among individuals of European ancestry, reflecting broader imbalances in genomic datasets. However, the key findings were consistent across multiple ancestral groups, supporting their generalizability.
“We’re not seeing fundamentally different genetic effects across populations,” Auton said. Still, increasing diversity in genetic research remains essential to ensure equitable advances in precision medicine. As digital tools continue to integrate genetic, clinical, and self-reported data, this participant-driven model may play an increasingly central role in biomedical discovery.
Putting pharmacogenomics in patients’ hands
Identifying genetic variants is only the first step, of course. The larger goal is to translate those discoveries into tools that can guide real-world decisions. To that end, the 23andMe Research Institute scientists developed predictive models that combine genetic information with clinical factors to estimate treatment outcomes.
The vision is straightforward: before starting a GLP-1 drug, a patient could receive a personalized profile indicating likely weight loss and risk of side effects. “People are making decisions about whether these medications are right for them,” Auton said. “Can we give them information to help with that decision?”
Such tools could have immediate clinical applications. A patient with a high predicted risk of nausea, for example, might start at a lower dose or follow a slower titration schedule. Another with a favorable genetic profile might be reassured about expected benefits.
For now, these findings are unlikely to immediately change prescribing practices, as clinical guidelines will require further validation through prospective studies. However, the trajectory is clear. In the near future, patients considering GLP-1 therapies may undergo genetic testing as part of routine care, with treatment decisions—such as drug choice, dosing, and expectations—guided in part by their DNA. For a class of drugs already transforming millions of lives, this approach could further enhance both efficacy and tolerability, underscoring that responses to GLP-1 therapies are shaped not only by pharmacology but also by the subtle variations of the human genome.
The broader significance of the study lies in its contribution to precision medicine: the idea that treatments should be tailored to individual biology rather than applied uniformly. In fields like oncology, this approach is already standard. But precision obesity treatment is in far earlier stages.
Auton is quick to re-emphasize that genetics is only one piece of the puzzle. Lifestyle, environment, treatment adherence, and underlying health conditions all shape outcomes. Still, even a partial predictive signal could be transformative in a field where trial-and-error prescribing is common.
As researchers continue to study GLP-1RAs, their potential appears to extend far beyond weight and blood sugar. Early evidence suggests benefits in cardiovascular health, inflammation, and even neurological conditions. Some studies are exploring their role in addiction and compulsive behaviors. “There’s an increasing literature that they’re beneficial in multiple areas,” Auton said.
This expanding scope makes understanding variability even more important. If GLP-1 drugs are to be used to treat a wide range of conditions, predicting who will benefit and who may be at risk becomes one of the most important, if not the most important, challenges.
What about sequencing?
Throughout our conversation, there was at least one elephant in the room. One is that this is not the first study to identify genetic variants influencing responses to GLP-1 drugs, as prior research has also implicated rs10305420. Slovenian researchers showed that genetic variability in GLP1R is associated with inter-individual differences in the weight-lowering-lowering potential of GLP-1 drugs in obese women with polycystic ovary syndrome (PCOS) in 2015, at a time when the main GLP-1 drug was liraglutide, which required daily injection.
More provocative is that the directionality of the variants’ effect reported in the Nature Medicine paper is the opposite of these previous studies. Auton’s team writes that such discrepancies may stem from differences in disease context, smaller sample sizes, limited statistical power, and variations in drug type, cohorts, and analytical methods.
Additionally, the GIPR variant rs1800437 (p.Glu354Gln) is already a known partial loss-of-function mutation, previously identified in a study of Chinese type 2 diabetes patients in 2019.
Perhaps the more significant issue is the question of sequencing. It’s not a space that 23andMe has completely avoided, as their premier consumer kit employs exome sequencing. But the cost of whole genome sequencing (WGS) direct-to-consumer products is now often priced lower than 23andMe’s premier kit, which goes for $499.
When asked about employing WGS, Auton revealed little of the calculus behind why 23andMe hasn’t added WGS to its arsenal of tools for interrogating genomes. “We’re very excited about that space,” Auton said. “Our focus has always been on what we can do in a direct consumer framework. There’s always been a price question there for WGS. It’s great. But when it was $1,000, it wasn’t obvious that that was going to be a compelling consumer offering. The pricing has reached its current level. It’s an area we’re very excited about and we’ll continue to look at.”
With studies like this, 23andMe 2.0 is making a case, perhaps its strongest yet, that its true value lies in something far more consequential: the ability to predict how individuals will respond to medicine before they ever take it. If that vision holds, the implications extend well beyond GLP-1 drugs. It suggests a future where prescribing a medication without first consulting a patient’s genetic profile feels incomplete, even irresponsible.
The post 23andMe Reports Genetic Predictors of Response to GLP-1 Drugs for Obesity appeared first on Inside Precision Medicine.
