Configuration Mimetic Osteogenic Peptides by Simulation and Machine Learning

Osteoporosis is a disease that weakens bones and makes them susceptible to fracture. It affects millions of people in the U.S., especially the elderly. Bone morphogenetic protein-2 (BMP-2) is a protein that can stimulate bone growth, enhance healing of damaged bone, and reduce pain in osteoporosis patients. However, its use is limited by complications such as bone overgrowth and tumor formation. An alternative is to use short amino acid chains called peptides that mimic the actions of BMP-2 without its complications. This project will explore how a peptide’s molecular conformation affects its biological activity. The project will employ machine learning to connect a peptide’s amino acids and molecular conformation with its ability to stimulate bone formation. The team will produce a database of peptides and their structures and then test them for therapeutic activity and bone regeneration. The outcomes will help identify effective peptides that can be used safely to treat damaged bone. The project will also develop an undergraduate course project and a workshop for high school teachers on machine learning in biotechnology. The hypothesis of this proposal is that the orders of magnitude lower osteogenic activity of the knuckle epitope peptide of bone morphogenetic-2 (BMP2-KEP) is rooted in configurational differences between its native state on the protein and its free state. The BMP2-KEP activity is much lower in the free state due to the collapsed state of the peptide structure. This hypothesis will be tested by building an ML-driven Quantitative Structure-Activity Relationships (QSARs) model for discovering new osteogenic peptides using a database of configurational properties of modified knuckle epitope of BMP2-KEP sequences. The sequences are predicted by mesoscale simulation followed by validation of the QSARs by experimental evaluation of the predicted peptides in a biomimetic tissue model to assess osteogenesis and immunogenicity. The project will also deliver the ML-predicted and experimentally tested peptides to mesenchymal stem cells safely and effectively by conjugation to nanogels. The outcomes of this project will advance knowledge of configuration-mimetic peptides and will help understand relations between peptide configuration and biological activity. The project will promote applications of therapeutic peptides that lead to clinical breakthroughs in regeneration of injured tissues and bone. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.