Fig. 1: fRIP-seq confirms XIST domains and reveals modular RNA–protein interactions.

a Schematic diagram of fRIP-seq experiment. Blue and red lines (R1 and R2) represent paired-end sequence tags. Gray lines represent the non-sequenced regions of RNA fragments, each ~200 nt. b fRIP-seq analysis strategy. Paired-end reads are mapped to the genome, which reveals non-sequenced fragment as a gap (gray line between R1 and R2). Mapped reads were remapped to the mature XIST. c Distribution of gaps between paired-end tags R1 and R2. Most tag-pairs are from one RNA fragment (left side). A small fraction of pairs are far from each other, therefore most likely from two proximally ligated fragments (right side, same distribution, but highlighting the long-distance pairs). One replicate was shown for each protein and the average read numbers and standard deviations are calculated from all biological replicates (n = 2 for EZH2 and n = 3 for the rest). d Human XIST RNA model, including exons, repeats, phylogenetic conservation (PhastCons 100 and Placental PhyloP from UCSC), and HEK293 PARIS data²⁷. Groups that correspond to four fRIP-seq LGs were extracted from PARIS data. e Long-distance arcs (tag pairs), coverage of long-distance tag pairs, and coverage of all tags for four examples as in c. f Comparison of the overlapping duplex groups from PARIS and the LGs from EZH2 fRIP-seq. g Comparison of PARIS and EZH2 fRIP-seq LG2 and LG2 extended to the average fRIP-seq fragment size (LG2extend [0–194]). Each side shows a 400-nt window. h Unsupervised clustering of XIST–protein interaction profiles in 100-nt windows. A total of 74 samples are clustered, excluding the samples STAG2 (non-specific, as determined by ref. ³⁵) and WDR5 (low read numbers). The bar graph on the right side shows the fraction of long-distance read pairs pulled down each protein relative to total mapped reads (see numbers in Supplementary Fig. 1a). n = 3 biologically independent samples. Data are presented as mean values +/− standard deviation in the bar graph. i PCA analysis of profiles in 100-nt windows. The numbers on the right are variation explained by each component. Source data are provided as a Source data file.