A0545
Title: Advancing graph neural networks for disease classification and feature selection in high-dimensional data
Authors: Tiantian Yang - University of Idaho (United States) [presenting]
Abstract: Omics data play crucial roles in exploring disease pathways, forecasting clinical outcomes, and gaining insights for disease classification. However, the significant challenge of dealing with a relatively small number of samples and a large number of features complicates the development of predictive models for omics data analysis. This challenge arises from inherent sparsity in biological networks and unknown feature interactions, adding further complexities. The advent of graph neural networks (GNN) helps alleviate these challenges by incorporating known functional relationships into a graph. However, many existing GNN models utilize graphs either from existing networks or generated ones alone, limiting model effectiveness. To overcome this restriction, an innovative GNN model is proposed that integrates information from both externally and internally generated feature graphs. The model is extensively tested through simulations and real data applications, confirming its superior performance in classification tasks compared to existing state-of-the-art baseline models. Furthermore, the GNN model can select features with meaningful interpretations in the biomedical context.
