Dottorato in DATA SCIENCE

Titolo della tesi: The Chromatin Network: A Modeling Strategy To Decipher Epigenetic Dynamics

Epigenetic dynamics are pivotal mechanisms supporting cellular identity, plasticity, and adaptation. In malignancy, the temporal modulation of epigenetic determinants enables cells to rapidly adapt in response to environmental or therapeutic pressure through epigenetic rewiring that supports transcriptional reprogramming. This plasticity can lead to disease progression and the development of drug resistance even in the absence of genetic alterations. Recent evidence supports the role of Cis-Regulatory Elements (CREs) in shaping cellular plasticity through the regulation of target genes' expression. Thus, understanding how CREs change their accessibility pattern over time, and how co-accessibility across these regions reflects coordinated regulatory activity, remains a central challenge. Although capturing the temporal relationship among CREs is essential to comprehend how cells respond to environmental changes, current computational approaches largely fail to model the dynamic coordination of regulatory programs over time. To fill this gap, we introduce T-ChroNet (Time-aware Chromatin Network), a network-based method that models cis-regulatory elements as nodes and their temporal co-accessibility as edges. T-ChroNet represents chromatin dynamics as a weighted undirected correlation network, enabling the identification of cis-regulatory elements sharing similar accessibility patterns over time. This approach facilitates the inference of putative upstream regulators and downstream biological pathways driving dynamic epigenetic changes. The application of T-ChroNet to a longitudinal dataset of Multiple Myeloma patients demonstrated its capability to integrate clinical features into the networks to answer biological questions. Indeed, through its application and the integration of multi-modal data, we defined ATF3 as a putative regulator of Bortezomib resistance and revelaed a subset of regulated genes that may be targeted to overcome the resistance problem in Multiple Myeloma patients treated with proteasome inhibitors. Overall, we presented a computational framework to analyze a longitudinal epigenetic dataset. Through the incorporation of the temporal variable, T-ChroNet advances our understanding of regulatory plasticity and provides a method to uncover epigenetic vulnerabilities in malignant contexts.