Slimming With Secretin: How Machine Learning Can Turn Secretin Into Potent GLP-1 Receptor Agonists
Background: GLP-1 (glucagon-like peptide-1) drugs are crucial in treating diabetes and obesity due to their ability to enhance insulin secretion, inhibit glucagon release, and slow gastric emptying, leading to better blood glucose control and reduced appetite. Recent research further suggests that GLP-1 receptor (GLP-1R) agonists may offer promising benefits in treating heart disease and Alzheimer's disease by reducing cardiovascular risk and potentially mitigating neurodegenerative processes. One major challenge for the synthesis and formulation of GLP-1 is its propensity to self-assemble into amyloid fibrils. Thus, the development of GLP-1 receptor agonists that do not oligomerize and are potent, selective, and long-acting has the potential to provide improved drug leads for the treatment of various diseases.
The August BPS journal club features an article by Nielsen et al. from Gubra (https://pubs.acs.org/doi/10.1021/acs.jmedchem.4c00417) that utilizes secretin as a backbone to create potent, selective, and long-acting GLP-1 receptor (GLP-1R) agonists. Secretin is a peptide hormone that, like GLP-1, belongs to the glucagon superfamily but, unlike GLP-1, does not aggregate. Thus, secretin was used as a backbone to create peptide analogs with improved physicochemical properties compared to GLP-1.
Summary of Findings: To identify improved GLP-1R agonists, Nielsen et al. developed a novel peptide drug discovery platform named streaMLine, which facilitates the large-scale design, synthesis, and screening of extensive peptide libraries, incorporating ML-driven quantitative structure-activity relationship (QSAR) analysis. Using this platform, the authors systematically explored secretin in an iterative manner, with several rounds of peptide design, synthesis, testing and ML-driven QSAR analysis. In total, the study resulted in the screening of 2,688 peptides, leading to the identification of multiple stable and potent GLP-1R agonists. One notable candidate, GUB021794, demonstrated significant in vivo efficacy by promoting body weight loss in diet-induced obese mice and exhibiting a half-life suitable for once-weekly dosing.
Potential Impact: The streaMLine platform exemplifies how integrating large-scale peptide synthesis and testing with ML can overcome traditional limitations, such as the scarcity of data for systematic analog design and the laborious nature of peptide synthesis and screening. The ability to generate and analyze large peptide libraries efficiently could accelerate the development of new therapeutics with improved properties, thereby enhancing the treatment options for diseases like diabetes and obesity. The specific success with GLP-1R agonists reported in this study highlights the platform's capability to refine and optimize peptide drugs, potentially leading to more effective and longer-lasting treatments.
Link to open access article: https://pubs.acs.org/doi/10.1021/acs.jmedchem.4c00417
Nielsen, J. C., Hjo Rringgaard, C., Nygaard, M. M. R., Wester, A., Elster, L., Porsgaard, T., Mikkelsen, R. B., Rasmussen, S., Madsen, A. N., Schlein, M., Vrang, N., Rigbolt, K. & Dalbo Ge, L. S. Machine-Learning-Guided Peptide Drug Discovery: Development of GLP-1 Receptor Agonists with Improved Drug Properties. J Med Chem, doi:10.1021/acs.jmedchem.4c00417 (2024).
Helena Safavi-Hemami, PhD
Member, BPF Scientific Advisory Board
linkedin.com/in/helena-safavi
Read previous editions of the BPF Journal Club series: https://www.boulderpeptide.org/journal-club