Extended Data Fig. 10: Properties of compartment repositioning events in post-WGD tumours and their consistency across different change-point detection algorithms.
From: Whole-genome doubling drives oncogenic loss of chromatin segregation
a, Distribution of compartment rank differences estimated by Calder (Δcompartment rank: negative values indicate a shift towards less active compartment, positive values indicate a shift towards a more active compartment) between RPETP53−/− control replicates (grey fill), WGD vs control (black line), and 20 weeks post-WGD tumours vs control (red line). b, Distribution of sub-compartment differences estimated by Calder (Δsub-compartment) in WGD and 20 weeks post-WGD tumours, compared to control. c, Schematic representation of the CoRE detection algorithm. The genome is partitioned in segments having the same Calder rank (blue representing Inactive, red representing Active regions). The differential rank values (Δr) between the tumour and the control samples were segmented using a recursive strategy. A null hypothesis of compartment repositioning from replicate comparisons (Δrrep) was derived and tested to compute the significance of each computed segment. The output of the algorithm is a list of activating and in-activating regions. d, Number of Mbp covered by activating and inactivating CoRE regions in the three 20 weeks post-WGD tumours. e, Distribution of CoRE region sizes (in bp) between activating and inactivating regions in the three 20 weeks post-WGD tumours. f, The relationship between differential compartment rank (x-axis) and differential ChIP-seq signal (y-axis) between the tumour and control sample for each tumour and each histone mark. Points represent 50-kb genomic bins. g, Expected distributions of correlations (grey) between changes in sub-compartment rank and changes in histone modification signal across different histone marks and different 20 weeks post-WGD tumours in RPETP53−/− cells. Observed correlations are highlighted as red lines. Empirical p-values are shown. p-values calculated by data permutation (n = 1000) h, Percentage of the genome covered by a CoRE in the three RPETP53−/− 20 weeks post-WGD derived tumour samples. i, Example of CoREs detected using the DiffComp or CBS algorithm for delta-rank segmentation. j, Distribution of distances between DiffComp and CBS breakpoints (red) and between DiffComp breakpoints and randomly generated ones by shuffling the CBS breakpoints across the genome. p-value calculated by Kolmogorov-Smirnov test. k, Volcano plot representation of CoREs detected by DiffComp (left) and CBS (right). CoREs involving the selected oncogenes and tumour suppressors discussed in the manuscript are labelled and detected by both approaches. p-values calculated by DiffComp. l, Number of activating and inactivating CoREs detected by both algorithm (shared) or by only one of them.
