Extended Data Fig. 5: Cancer-specific allele normalization and neo-TAD identification.
a–c, We show an example here in SK-N-MC cells. As shown in a, there is a 1.5 Mb deletion event (chr8: 127.88M, +; chr8: 129.37M, -). Since this is a heterozygous deletion, the chromatin interactions in area A3 are between loci (127.28M – 127.88M) and loci (129.37M – 129.97M) on one allele where the deletion happens. On the contrary, chromatin interactions within A1 (or A2) area are from all alleles and therefore, the overall intensities in area A3 are much weaker than A1 or A2. We need to normalize the signals so that we can predict neo-TADs and neo-Loops. a, Hi-C matrix without normalizing cancer-specific allele. The neo-TAD is undetectable as shown by the directionality index (DI) track. b, Reconstructed Hi-C map after cancer-specific allele normalization. Now the neo-TAD becomes detectable, and the number of detected neo-loops is also enhanced. c, Linear regression of the local distance averaged contact frequencies and the global distance averaged contact frequencies in different regions (A1, A2 and A3) of the Hi-C map in a. d,e, Detection of neo-TADs in 50 cancer cell lines or patient samples. d, The number of neo-TADs detected in each sample. e, Aggregate analysis of neo-TADs and distribution of breakpoint locations. Hi-C signals were distance-normalized, averaged, and centered at neo-TAD midpoints.
