When a patient on a glucagon-like peptide-1 (GLP-1) receptor agonist loses a lot of weight, hundreds of plasma proteins change. The challenge is working out which changes come from the drug itself and which simply follow the weight loss. Only the first kind can tell you something about how the drug works.
At Lilly, I apply large-scale proteomics (mainly Olink Explore HT and SomaScan 11k) to phase 2 and phase 3 trials in obesity and cardiometabolic disease. I use longitudinal mixed models to track proteins over time, mediation analysis to test whether protein changes explain clinical outcomes, dose-response comparisons across treatment arms, and genetic methods (Mendelian randomization, protein quantitative trait locus (pQTL) colocalization) to separate causal signals from confounding. I also build the pipelines, data models, and dashboards the clinical omics team works with.
SURMOUNT-5: tirzepatide versus semaglutide
SURMOUNT-5 compared tirzepatide (10 and 15 mg) head-to-head with semaglutide (1.7 and 2.4 mg) in adults with obesity. Tirzepatide led to more weight loss, so comparing the arms means separating protein changes that track weight loss from those that don’t. In exploratory analyses, tirzepatide was associated with different changes in the plasma proteome than semaglutide. I presented this work in an oral session at the European Association for the Study of Diabetes (EASD) annual meeting, EASD 2026 in Milan.
The analysis code is public on GitHub at EliLillyCo/surmount5-proteomics, and an interactive results dashboard is available at surmount5-proteomics.lilly.com.
Related: UK Biobank Pharma Proteomics Project, the population proteomics resource behind some of the methods described above.