Multi-omics and machine learning approaches for complex diseases

Complex diseases arise from an intricate interplay of genetic, environmental, and lifestyle factors, making them difficult to decipher. Integrating multiple molecular layers such as genomics, epigenomics, transcriptomics, and proteomics enables a more comprehensive understanding of disease mechanisms. While this multi-omics perspective offers unprecedented opportunities, it also introduces major challenges in data integration and interpretation. In this context, machine learning (ML) represents a powerful tool to analyze heterogeneous datasets, uncover hidden patterns, and prioritize features most relevant for diagnosis, prognosis, and therapeutic development. This thesis is structured around three projects aimed at exploring how multi-omics data and machine learning can be combined to advance the study of complex diseases. In the first project, we investigated stochastic epigenetic mutations (SEMs) as biomarkers of disease reactivation in Juvenile Idiopathic Arthritis (JIA). We demonstrated that SEM burden can both reflect and precede clinical reactivation among patients with otherwise indistinguishable clinical profiles. Furthermore, by adopting a multi-omics approach, we identified candidate epigenetically-driven differentially expressed genes and prioritized potential therapeutic targets. In the second project, we developed machine learning models to predict cell type-specific regulatory variants contributing to Alzheimer’s disease heritability. These models enabled the identification of novel candidate expression Quantitative Trait Loci (eQTLs) and risk genes, offering new perspectives on the molecular mechanisms underlying the disease. In the third project, we investigated Syndrome of Undifferentiated Recurrent Fever (SURF) through a hierarchical integration of genome-wide analyses and multi-omics QTL approaches. This integrative strategy enabled the identification of candidate molecular mechanisms linking genetic variation to inflammatory pathways and refining disease-specific biomarkers.