The School of Computer Science would like to present…
Cell Type Prediction in Heterophily Graphs Using Graph Neural Networks
PhD Comprehensive Exam by: Mahshad Hashemi
Date: Tuesday, 23 July 2024
Time: 11am
Location: Room 122, Essex Hall
Abstract:
Accurately predicting cell types from single-cell RNA sequencing (scRNA-seq) data is a critical objective in computational biology, essential for understanding cellular diversity and function. This study addresses this challenge by employing Graph Neural Networks (GNNs) within the framework of heterophily graphs, characterized by connections between nodes with differing labels or features. Traditional GNNs are optimized for homophilic graphs, where similar nodes are interconnected, limiting their performance in heterophilic biological networks. Our research investigates the capability of GNNs to effectively manage heterophily, aiming to enhance the precision and reliability of cell type predictions in scRNA-seq data. By leveraging the distinctive properties of heterophilic networks, we strive to improve cell type classification and advance our comprehension of complex biological systems.
In this presentation, I will delve into four critical topics that underpin our research. First, I will explore the principles and applications of Graph Neural Networks, highlighting their relevance in biological data analysis. Second, I will discuss the concepts of heterophily and homophily, explaining how these structures influence network connections and affect GNN performance. Third, I will examine cell-cell interactions, emphasizing their role in cellular communication and network formation. Finally, I will address ligand-receptor interactions, providing insights into how these interactions contribute to our understanding of cell type prediction in scRNA-seq data. These topics collectively provide a comprehensive understanding of the methodologies and insights driving our approach to cell type prediction.
Keywords:
Graph Neural Network, Heterophily, Homophily, Single Cell RNA Sequencing, Cell Type Prediction
PhD Doctoral Committee:
External Reader: Dr. Mitra Mirhassani
Internal Reader: Dr. Pooya Moradian Zadeh
Internal Reader: Dr. Jianguo Lu
Advisor(s): Dr. Luis Rueda and Dr. Alioune Ngom
MAC STUDENTS ONLY - Register here