Phenomix AI algorithm can predict GLP-1 side effects

Phenomix Sciences, working in partnership with Mayo Clinic researchers, has presented new data showing its proprietary machine-learning algorithms can predict which patients are more likely to experience side effects, specifically nausea, from GLP-1 therapies.

Phenomix Sciences' MyPhenome test identifies obesity phenotypes, including Hungry Gut (accelerated gastric emptying), Hungry Brain (a person who consumes too many calories without feeling full), and Emotional Hunger (individuals who eat in response to emotional triggers).

The research, presented at findings at Digestive Disease Week 2025, was led by renowned obesity researcher and Phenomix co-founder, Dr Andres Acosta and presented by Dr Thomas Fredrick, marks an important step toward more personalised obesity care and improved clinical trial design to support drug development.

The study titled, ‘A Genetic Risk Score Associated with Nausea Resulting from GLP-1 Agonist Treatment: A Post-Hoc Analysis of a Randomized Controlled Trial of Liraglutide’, analysed post-hoc genetic data from 110 participants. Using a machine learning-assisted Genetic Risk Score (GRS), researchers analysed the relationship between individual genetic profiles and adverse events, including nausea. Patients with a high GRS were more than twice as likely to experience nausea from liraglutide, a GLP-1 medication, compared to those with a low score (68% vs. 30%).

Nausea is the most common side effect of GLP-1s like liraglutide, with up to 40% of patients experiencing it and up to 6.4% stopping treatment because of it. This kind of predictive insight helps reduce waste, prevent avoidable ER visits, and ensure patients are matched with medications they can tolerate from the start. It also has implications in pharmaceutical development, improving participant selection and retention in clinical trials, and accelerating time to market.

"These findings represent a meaningful advancement in how we approach obesity treatment at an individual level," said Acosta. "By identifying which patients are more likely to experience side effects before starting therapy, we can improve tolerability, support long-term adherence, and better match the right treatment to the right patient. This is a critical step toward delivering on the promise of truly personalized obesity care."

Previous research presented by Phenomix at DDW 2024 demonstrated that Phenomix Sciences' MyPhenome test can identify patients more likely to respond to semaglutide. MyPhenome is a saliva swab that determines the root biological factors that can cause obesity and helps physicians personalise treatments for more effective weight loss. This year's findings zero in on the patients who may see optimal weight loss but still experience treatment-limiting side effects.

"Our team's research builds on previous findings by showing we can now predict not just who will benefit from GLP-1s, but who is more likely to struggle with side effects," said Fredrick. "That allows for more balanced, individualised treatment planning. It's an important advancement in the clinical application of precision obesity medicine."

"This study underscores the power of predictive tools like MyPhenome to transform how we approach obesity treatment, not just in the clinic, but in the drug development pipeline," said Mark Bagnall, CEO of Phenomix Sciences. "By identifying patients at risk for side effects before treatment begins, we can match the right patient to the right therapy, increase real-world adherence, and dramatically improve clinical trial efficiency through smarter patient selection. Our strategic partnership with Mayo Clinic, and its dedicated research team including Drs Acosta and Frederick, have been critical in validating this precision medicine approach."

The study was one of 17 presented by Mayo Clinic researchers, with eight incorporating Phenomix's machine learning-based algorithms. Together, the research highlights the growing role of precision medicine in advancing obesity care and drug development.