Abstract B009: Computational discovery of dynamic protein folding landscapes reveals novel osteogenic growth factor mimetics for craniofacial bone regeneration

Abstract

Regeneration of craniofacial bone defects remains limited by the high cost, short biological half-life, and adverse effects associated with recombinant osteogenic growth factors such as BMP-2. Most computational protein design approaches optimize existing protein scaffolds rather than discovering entirely new structural motifs capable of activating osteogenic signaling. We developed a computational protein engineering framework integrating de novo protein design, structural landscape modeling, receptor interface prediction, and molecular docking to identify novel protein folds predicted to function as osteogenic growth factor mimetics. A library of 100,000 de novo protein scaffolds was computationally generated and evaluated for structural stability using protein folding simulations. Candidate structures were screened against 250 receptor interface conformations representing osteogenic signaling proteins involved in BMP, TGF-β, and related regenerative pathways. Approximately 2 million protein-receptor structural comparisons were performed using geometric interface alignment, molecular docking, and binding energy estimation to identify scaffolds capable of engaging osteogenic receptors through previously uncharacterized interaction geometries. Structural clustering and conformational landscape analysis were used to prioritize diverse folding motifs while eliminating redundant scaffold architectures. The computational pipeline identified 428 structurally stable scaffolds exhibiting favorable folding energetics and receptor accessibility. Of these, 36 candidates demonstrated consistently high predicted binding affinity across multiple osteogenic receptor conformations, while 11 de novo folding motifs exhibited receptor interaction geometries distinct from known BMP family proteins. Network-based pathway analysis predicted that these candidates preferentially engage signaling modules associated with osteoblast differentiation, extracellular matrix organization, and skeletal development. Compared with optimization of existing BMP-derived scaffolds, the de novo design framework increased recovery of high-affinity candidate mimetics by 38% while substantially expanding structural diversity across prioritized proteins. These findings demonstrate that exploration of dynamic protein folding landscapes enables discovery of previously unrecognized osteogenic protein architectures beyond naturally occurring growth factor families. By integrating large-scale de novo protein design with structural bioinformatics and receptor interface modeling, this computational framework provides a scalable strategy for identifying next-generation regenerative protein therapeutics for craniofacial bone repair. The approach establishes a foundation for experimental validation of synthetic osteogenic growth factor mimetics and may accelerate development of safer, more durable biologics for skeletal tissue engineering and regenerative medicine. Shivi Kumar, Philip Moheno, Jennifer Lu. Computational discovery of dynamic protein folding landscapes reveals novel osteogenic growth factor mimetics for craniofacial bone regeneration [abstract]. In: Proceedings of the 19th AACR Conference on the Science of Cancer Health Disparities; 2026 Oct 18-21; Atlanta, GA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2026;35(10 Suppl):Abstract nr B009.