FuDiCo: Gene Fusion-Initiated Path Propagation for Disease Comorbidity Prediction

Disease comorbidity—the co-occurrence of two or more diseases in the same individual—has gained growing attention due to its association with adverse clinical outcomes and increased treatment complexity. Recent subgraph-based approaches for disease comorbidity prediction model disease modules as subgraphs induced by disease-associated genes in the protein–protein interaction (PPI) network and learn disease representations from subgraph topology. However, these approaches are constrained by incomplete disease–gene annotations, which may obscure important molecular relationships between diseases. Accordingly, disease comorbidity may also be influenced by molecular events beyond annotated disease genes, such as gene fusion events that have emerged as important contributors to disease mechanisms. Motivated by the role of gene fusions in disease development, we introduce Gene Fusion-Initiated Path Propagation for Disease Comorbidity Prediction (FuDiCo), a framework that models comorbidity through influence propagation over the PPI network. FuDiCo represents fusion-associated genes as localized perturbation sources and learns how their influence propagates along interaction paths toward disease subgraphs, thereby capturing propagation patterns that link related diseases and contribute to their comorbidity. Experiments on a benchmark disease comorbidity dataset show that FuDiCo outperforms state-of-the-art methods, achieving statistically significant improvements. These results shed light on the importance of gene fusion events in understanding disease relationships.