Genome conformation and transcription regulation: methods and applications

The 3D organization of chromatin within the nucleus is crucial for genome functionality. This is true at multiple levels of resolution: on a large scale, with chromosomes occupying distinct volumes (chromosome territories), at the level of \nindividual chromatin fibers, organized in compartmentalized domains (as the Topologically Associating Domains, TADs), and down to the formation of short range chromatin interactions (as enhancer-promoter loops). The widespread adoption of high-throughput techniques derived from Chromosome Conformation Capture (3C) has been instrumental in advancing the knowledge of chromatin nuclear organization. In particular, Hi-C has the potential to achieve the most comprehensive characterization of chromatin 3D interactions, as in principle it can detect any pair of restriction fragments connected as a result of ligation by proximity. The analysis of the enormous amount of genomic data produced by Hi-C required the development of ad hoc algorithms and computational procedures. Despite the increasing number of available bioinformatics pipelines, no consensus on the optimal approach to analyze Hi-C data has been reached yet. Therefore, we quantitatively compared several Hi-C data analysis methods for the identification of multi-scale chromatin structures to highlight strengths and weaknesses of the various methods and propose application guidelines and good practices. Specifically, we compared different computational approaches (6 for the characterization of chromatin loops and 7 to identify TADs) using publicly available Hi-C datasets, comprising data from different species and cell lines, Hi-C protocol variations and data resolution. Additionally, the algorithms were tested on simulated Hi-C data to assess sensitivity and precision of each method. The tools differed in terms of implemented analysis steps and strategies adopted for alignment, filtering, normalization, and feature identification (global or local looping interactions calling and single-scale or multi-scale TAD discovery). Results of this comparison indicate that performances of the methods considerably vary, both in quantitative and qualitative terms, and that the tools need extensive optimization of the parameters in order to work properly. Despite, in general, TAD callers resulted riper than algorithms to call interactions, still most of them are characterized by crucial limitations, as for instance the inability to investigate how the 3D organization of chromatin structures \nevolves over time (as e.g., during differentiation). Although the molecular mechanisms underlying TADs formation are still debated, it is evident that distinct interaction patterns can be observed within individual TADs. In particular, some domains appear to have a very compact structure, while others have a less uniform or weaker interaction frequency within the domain, while showing a strong interaction between the borders. \nTo address these limitations, I developed TAD-AH (TADs Advanced Hierarchy), a four-step sequential procedure coded in R, for the characterization of both static and dynamically changing chromatin domains. As a case study, I analyzed Hi-C data generated prior and post human fibroblasts (IMR90) trans-differentiation into skeletal muscle cells (myoblasts, and, when put in differentiation media, myotubes) by overexpression of muscle stem cells master regulator MyoD. I integrated Hi-C with epigenomic and transcriptomic data from the same conditions and confirmed that the identified genomic features are consistent with the biological scenario under scrutiny.