Ligand-Receptor Dynamics in Heterophily-Aware Graph Neural Networks for Enhanced Cell Type Prediction - PhD Seminar by: Mahshad Hashemi

The School of Computer Science at the University of Windsor is pleased to present …

 

Date: Monday, May 26^th, 2025

Time: 11:00 am

Location: Essex Hall, Room 122

 

This seminar presents our study on improving cell type prediction from single-cell RNA-seq data using heterophily-aware Graph Neural Networks (GNNs). Unlike traditional GNNs that assume similar cells interact, we focus on biologically realistic scenarios where dissimilar cell types communicate through ligand-receptor (L-R) signals. Using LIANA, we construct L-R-based cell graphs and evaluate several GNNs—including GCN, H2GCN, and GBK-GNN—across six datasets. Our results show that GBK-GNN consistently outperforms others in low-homophily settings. This work demonstrates the importance of modelling heterophily in biological networks for accurate and interpretable cell type prediction.

 

Internal Reader:  Dr. Pooya Moradian Zadeh

Internal Reader: Dr. Jianguo Lu

External Reader: Dr. Mitra Mirhassani

Advisor(s): Dr. Luis Rueda, Dr. Alioune Ngom

 

Registration Link  (Only MAC students need to pre-register)