This thesis explores how bioinformatics advances the study of complex diseases by providing system-level models that capture intricate gene-protein interactions. Traditional reductionist methods focusing on isolated genes or proteins fail to explain the broader dysfunctions of complex diseases. The rise of diseases like cancer, cardiovascular and neurological disorders, and autoimmune conditions underscores the need for approaches leveraging high-throughput technologies and network-based models for comprehensive understanding.
Central to this research are functional association networks, which map direct and indirect functional relationships between genes and proteins. These networks integrate genomic, transcriptomic, proteomic, and functional data. Functional association networks are valuable resources in biological and biomedical research, enabling e.g. discovery of interaction partners, pathway enrichment analysis, and disease module identification. This thesis presents the 5th and 6th releases of FunCoup, a comprehensive resource of functional association networks, with upgrades in data integration, network framework, and user access via a redesigned website and Cytoscape app.
FunCoup 5 expanded to 22 species and generated around 70 million interaction links from 11 evidence types. Innovations included directed regulatory links, a SARS-CoV-2 virus-host network, and tissue-specific filtering for context-sensitive research.
FunCoup 6 expanded to all 640 species in InParanoiDB, with over 100 million links for the 23 primary species, including 1 million directed transcription factor-target links. A key improvement is the shift to a bin-free Bayesian training framework, using kernel density estimation of likelihoods of interaction from ten evidence types. New tools like MaxLink and TOPAS were added for disease module and drug target identification, together with built-in KEGG pathway enrichment via ANUBIX and EASE. Rebuilt in Python with a redesigned website, FunCoup 6 offers enhanced flexibility, scalability, and user access.
The FunCoup Cytoscape app uses the RESTful API to integrate FunCoup’s resources into Cytoscape, enabling network visualization and analysis across all 640 species with powerful Cytoscape tools and plugins.
This thesis highlights applications in pathway enrichment analysis (EA) and disease module detection. Network-based EA tools simplify the analysis of differentially expressed genes by linking them to known pathways, providing a systems-level understanding of molecular changes. A new benchmark compared 14 EA methods across 82 datasets from 26 diseases, showing that network-based methods like ANUBIX outperformed traditional approaches, offering more accurate insights and guidance for method selection.
TOPAS, a new approach, uncovers complex molecular interactions using functional association networks and disease-gene association databases. Complex diseases arise from genetic and environmental factors creating clusters of dysfunction within networks. These "disease modules" reveal genetic interactions in disease phenotypes and help identify novel therapeutic targets. By clustering disease-related proteins, this work advances network medicine, showcasing the power of computational techniques in understanding complex diseases.
This thesis contributes to network biology and network medicine, advancing multi-omics data integration into functional association networks, with a focus on FunCoup and network-based applications.