A Multi-Omics Classification Framework using Associative Graph Neural Networks with Prior Knowledge for Biomarker Identification

This thesis presents a new way to find important biomarkers linked to cancer by combining multiple types of biological data and expert knowledge. It uses a special type of artificial intelligence called graph neural networks to model complex relationships between biomarkers. The method was tested on breast and kidney cancer data and showed better accuracy than current tools. It also helped identify promising biomarkers for diagnosis. This research may lead to better cancer detection, more targeted treatments, and improved patient care.