SUN-213 Leveraging Unexplored Chemical Space and Computational Methods to Design the Next Generation of Glucagon Receptor Antagonists

Abstract Disclosure: J.M. McDaniel: None. M.E. Bramlett: None. A.S. Walker: None. Glucagon receptor (GCGR) antagonists offer a potential avenue to one vital characteristic not currently possessed by any drug approved to treat diabetes: beta-cell regeneration. Recent research has shown the capacity of GCGR antagonists to not only improve glycemic control by reducing excess glucose production, but also to restore beta-cell mass. Despite this enormous potential, serious adverse effects, including alpha-cell hyperplasia, hyperglucagonemia, and increased risk of hypoglycemia, have hindered the development of these drugs. In order to produce GCGR antagonists that are actually suitable for human disease treatment, it may be necessary to embrace previously unexplored and innovative approaches. Tapping into this potential, we explore here a novel approach to GCGR antagonist design, leveraging both unexplored chemical space and cutting-edge computational methods. Existing GCGR antagonists include numerous small molecules, glucagon-derived peptides, and monoclonal antibodies, but ignore the vast amount of chemical space between them. To take advantage of this space, researchers have developed several scaffolds for producing antibody mimetics and other small protein binders. These scaffolds can help hold binding sites in the correct geometry for optimal function, while also decreasing drug half-life and opening the door to easier chemical modification. Using computational modeling and machine learning approaches, we have designed antibody mimetic binders derived from known GCGR interactions. The Rosetta Residue Energy Breakdown application was used to identify anchor hotspot interactions between the extracellular domain of the GCGR and a known GCGR antagonist antibody, mAb1. Using Rosetta Anchored Design, these known interactions were then grafted onto an extensively validated scaffold, the Fibronectin Type III domain. The top designs were optimized using ProteinMPNN, a deep learning-based protein sequence design method. This presentation will include an overview of the design workflow, highlight the top candidates, and share preliminary experimental data evaluating their potential. Presentation: Sunday, July 13, 2025