Exploring disease-gene interactions with a multiplex heterogeneous network model

Network medicine has emerged as a promising approach for understanding biological complexity and gaining new insights into disease mechanisms. This thesis investigates the use of multilayer networks, and particularly a multiplex-heterogeneous network model, to model disease landscapes and relationships. Using a random walk with restart to explore disease-disease interactions, the effectiveness of various network models in reflecting biological realities was assessed. Furthermore, we compared the quality of predictions by larger, combined models, like a multiplex-heterogeneous network, to smaller ones. Our findings show that the network models are able to successfully distinguish between diseases based on genetic and symptomatic similarity. However, nuanced differentiation between diseases with only partial similarities remains challenging. This can likely be improved by enhancing the diversity as well as quality of underlying data. Pathway-enrichment analysis confirmed the ability of all network models to identify disease-related genes. Notably, random walk with restart on the multiplex-heterogeneous network showed particular promise in leading to the most disease relevant insights. These results underscore the potential of network-based methods to uncover novel disease insights, but also suggest that improvements in data quality can further enhance predictive accuracy.