A Graph Learning Framework To Identify Shared Genetic Drivers of CHD and Neuroblastoma

Monday, May 18, 2026

11:00 AM-1:00 PM

BIOMED PhD Research Proposal 

Title: 
A Graph Learning Framework To Identify Shared Genetic Drivers of Congenital Heart Defects (CHD) and Neuroblastoma

Speaker:
Benjamin Stear, PhD Candidate
School of Biomedical Engineering, Science and Health Systems
Drexel University

Advisors:
Deanne Taylor, PhD
Research Associate Professor of Pediatrics
Department of Biomedical and Health Informatics
Perelman School of Medicine
University of Pennsylvania 
Director, Bioinformatics Group
Children's Hospital of Philadelphia (CHOP)

Ahmet Sacan, PhD
Teaching Professor
School of Biomedical Engineering, Science and Health Systems
Drexel University 

Details:
Congenital heart defects (CHDs) and neuroblastoma (NBL) are major contributors to pediatric morbidity and mortality, yet their genetic etiologies remain poorly understood. Converging evidence suggests a shared developmental origin rooted in neural crest biology, in which perturbations to early cell fate decisions and regulatory programs may give rise to both structural birth defects and pediatric malignancy. However, conventional approaches such as single-trait genome-wide association studies are limited in their ability to capture the complex, pleiotropic, and network-driven nature of these diseases.

This work develops an integrative framework to identify and characterize shared genetic mechanisms underlying CHD and NBL by combining whole-genome sequencing data with biological interaction networks and single-cell transcriptomic trajectories. Graph neural networks are used to prioritize disease-associated variants and genes within an interactome context, while network propagation and enrichment analyses map genetic burden onto neural crest gene regulatory networks to identify functionally relevant modules. To resolve temporal and cell state–specific effects, single-cell trajectory data are incorporated into a state-resolved graph, enabling the application of temporal graph attention models to infer how genetic perturbations influence developmental progression.

Together, this approach provides a systems-level view of pleiotropy by linking genetic variation to dynamic regulatory processes across development. The results are expected to uncover previously unrecognized disease-associated genes, identify critical developmental windows of vulnerability, and provide mechanistic insight into how disruptions in neural crest development give rise to divergent pediatric phenotypes. This framework offers a generalizable strategy for studying complex developmental diseases and advancing precision medicine in pediatric populations.

Contact Information

Natalia Broz
njb33@drexel.edu