TADfit is a multivariate linear regression model for profiling hierarchical chromatin domains on replicate Hi-C data

Abstract

Topologically associating domains (TADs) are fundamental building blocks of three dimensional genome, and organized into complex hierarchies. Identifying hierarchical TADs on Hi-C data helps to understand the relationship between genome architectures and gene regulation. Herein we propose TADfit, a multivariate linear regression model for profiling hierarchical chromatin domains, which tries to fit the interaction frequencies in Hi-C contact matrix with and without replicates using all-possible hierarchical TADs, and the significant ones can be determined by the regression coefficients obtained with the help of an online learning solver called Follow-The-Regularized-Leader (FTRL). Beyond the existing methods, TADfit has an ability to handle multiple contact matrix replicates and find partially overlapping TADs on them, which helps to find the comprehensive underlying TADs across replicates from different experiments. The comparative results tell that TADfit has better accuracy and reproducibility, and the hierarchical TADs called by it exhibit a reasonable biological relevance.

Identifying TAD-like domains on single-cell Hi-C data by graph embedding and changepoint detection

“…This interesting observation encourages us to further explore the biological meaning underlying the enrichment. In light of some reports that higher enrichment of architectural proteins and histone marks is usually accompanied by higher levels of hierarchical TADs on bulk Hi-C data ( Wang et al 2017 , An et al 2019 , Cresswell et al 2020 , Liu et al 2022 ), we speculate that there might be an association between TAD-like domains on single-cell Hi-C data and hierarchical TADs on bulk Hi-C data. To check this point, the hierarchical TADs were called with a recently developed tool named TADfit ( Liu et al, 2022 ), which is a multivariate regression model for the identification of hierarchical TADs on replicate Hi-C data.…”

“…In light of some reports that higher enrichment of architectural proteins and histone marks is usually accompanied by higher levels of hierarchical TADs on bulk Hi-C data ( Wang et al 2017 , An et al 2019 , Cresswell et al 2020 , Liu et al 2022 ), we speculate that there might be an association between TAD-like domains on single-cell Hi-C data and hierarchical TADs on bulk Hi-C data. To check this point, the hierarchical TADs were called with a recently developed tool named TADfit ( Liu et al, 2022 ), which is a multivariate regression model for the identification of hierarchical TADs on replicate Hi-C data. And the hierarchical level of a bin can be assigned using the terminology given by An et al (2019) , where the boundary level of a bin is defined as the maximum number of hierarchical TADs that use a boundary on either its left or right side.…”

“…Then, on both the full and downsampled bulk Hi-C data, scKTLD was compared with the other seven TAD callers, including deTOKI, deDoc, scHiCluster, TopDom, SpectralTAD, GRiNCH, and Higashi. It is worth noting that among these methods, deTOKI, scHiCluster, and Higashi are devoted to the identification of TAD-like domains on single-cell Hi-C data, while the remaining four are designed for bulk Hi-C data and are reported to be applicable to lower sequencing depths ( Cresswell et al 2020 , Lee and Roy 2021 , Liu et al 2022 ). This comparison was conducted in the following three ways: Firstly, similarity between the TADs was called for on bulk Hi-C data before and after downsampling.…”

“…The boundaries of TADs are reported to be conserved across cell types and even species, demarcated by CTCF, housekeeping genes, cohesin complexes, and other histone marks ( Dixon et al 2012 , Bonev and Cavalli 2016 ), and the disruption of TADs may result in gene misregulation and severe diseases, such as developmental disorders and cancers ( Lupiáñez et al 2015 , Flavahan et al 2016 , Hnisz et al 2016 , Lupiáñez et al 2016 ). Besides, further studies show that TADs are organized into a hierarchy with different structural levels ( Weinreb and Raphael 2016 ), the more TADs move towards the inner of hierarchy, the higher their levels, and there is a more remarkable CTCF enrichment and gene expression near the boundaries of TADs with higher levels ( An et al 2019 , Cresswell et al 2020 , Liu et al 2022 ).…”

Efficient Hi-C inversion facilitates chromatin folding mechanism discovery and structure prediction

“…Analytical techniques that use Hi-C data have expanded our understanding of the regulation of chromatin loops themselves by identifying the location of loop anchors and relating them to sequence information. However, identifying these interactions often requires requires sophisticated algorithms [7,[67][68][69][70][71][72], as the aforementioned influence of polymer topology complicates the relationship between the functional importance of an interaction and its probability. Because our algorithm separates these effects from the influence of local interactions, we anticipate that contact energy extraction will facilitate the identification of functionally important interactions.…”