Please join us for a UW Data Science Seminar featuring UW Bioengineering PhD Student Ike Keku on Thursday, January 22nd from 4:30 to 5:20 p.m. PT. The seminar will be held in IEB G109.
“Using AI/ML Models to Predict Cross-Family Ligand Binding to VEGFR”
Abstract: Angiogenesis, the formation of new blood vessels, is tightly regulated by vascular endothelial growth factor receptors (VEGFRs). This process is fundamental to both physiological processes (e.g., wound healing, tissue regeneration) and disease states (e.g., cancer progression, diabetic retinopathy, atherosclerosis). Despite clinical success, anti-VEGF therapies face resistance and limited durability. This limitation is partly due to the fact that angiogenic signaling involves more than a single ligand–receptor family. In addition to VEGF, platelet-derived growth factors (PDGFs) can also bind to VEGFRs. These interactions may affect endothelial cell behavior and treatment effectiveness. Still, these cross-family interactions are not fully understood. Because experimental protein-protein interaction (PPI) determination is time-consuming and costly, computational approaches hold promise for large-scale interactome studies.
Evaluating PPIs with various sequence-structural machine learning (ML) classifier combinations could help uncover proteins that interact across families, with the advantage of greater availability and accessibility compared to sequence data or structural data alone. Existing models frequently lack standardized evaluation measures and instruments for comparison. To close this gap, we propose PPInsight, a Python-based benchmarking and visualization tool that automates the extraction of reference interaction data, runs various predictive models, and produces comparative performance graphs. This approach will allow researchers to easily evaluate model correctness and consistency. Through this, the predictive performance of predictor models for PPIs between VEGFRs and cross-family ligands can be evaluated, guiding experimental validation efforts and the downstream development of simulated vascularization models. By understanding the computational boundaries and strengths of different model types, this work may improve the prediction of uncharacterized ligand–receptor interactions and offer advancements in targeted therapy development for angiogenesis-dependent diseases.
Speaker Bio: Ike Keku is a Bioengineering PhD Student at the University of Washington. He works in the Research Lab of Dr. Princess Imoukhuede, where he studies systems biology, data science, and protein-protein interactions. He completed his undergraduate work at the University of North Carolina at Chapel Hill, earning a BS in Biomedical Engineering.