Fig. 7. The caRNA-gene interactome preferentially links upregulated caRNAs to upregulated genes.

a Representation of the caRNA-gene interactome as a matrix containing the number of contacts between an ncRNA (row) and the proximal regulatory region (PRR) of a protein-coding gene (column). Only cis interactions are shown for simplicity. b caRNA-gene interactome in ES and DE cells for the 50 most abundant lncRNAs (top) and UTLs (bottom) on Chr11 (left). Expanded view of the interactome for 50 protein-coding genes upstream and downstream of each caRNA PRR (right). Expanded maps are shown for the true interactome signal (Obs), the generative model prediction (Mod), the log2 fold change of the observed over model (Obs/Model), and the interactions significantly enriched in the observed over the model (Sig, p-value < 0.05 and LFCobs, model > 0)as described in Fig. 5c, d (“Methods”). c Volcano plot showing the differential lncRNA-gene contacts is ES versus DE cells. Each data point is a contact between a lncRNA and the PRR of a protein-coding gene. log2 Fold Change of contact rate in DE versus ES cells (LFC_ES, DE) and False Discovery Rate adjusted p-values were computed with DESeq2 as in Fig. 3a, and colored contacts are those with an adjusted p-value < 0.05. d Quantification of the percentage of cell-state-specific contacts for each class of caRNA relative to the number of contacts tested for that class (top) and number of distinct caRNAs involved in these contacts (bottom). Cell-specific contacts were defined as those with an adjusted p-value < 0.05 and LFCES,DE > 1.3 by DESeq2. e Top 20 lncRNA-gene contacts upregulated in ES (left) and DE cells (right) in the observed data (blue circles). Most of these contacts are also predicted to be among the 20 most upregulated contacts by the generative model (purple circles). f Scatter plots showing for each differential contact the relationship between the change in contact rate during differentiation (LFCES,DE) and the change in the chromatin levels of the involved caRNA (left) and in the expression of the cognate protein-coding gene (right). Differential contacts were defined as in (d). Only differential contacts involving exons of lncRNAs or UTLs are shown. g Percent of protein-coding genes targeted by one or more dynamic contact with a lncRNA (left panel), a CRE-derived RNA (middle panel), or any UTL (right panel, excluding tRNA- and snRNA-derived NARs). Protein-coding genes are grouped (x-axis) according to whether their expression is upregulated in ES, DE, or stable during differentiation as measured by total RNA-seq (DEseq2, FDR 0.05, fold change threshold 3x). Colors indicate whether the protein-coding gene is targeted by a single (light colors) or several (dark colors) caRNAs with which the interaction is upregulated in ES (blue shade) or DE (yellow shade). Some genes are targeted by several caRNAs, which include both ES and DE upregulated interactions (purple). h Top two rows: Percentage of interactions upregulated in ES targeting a protein-coding gene upregulated in ES (positive association) or targeting a protein-coding gene upregulated in DE (negative association). Bottom two rows: similarly, for interactions upregulated in DE cells. i Fold enrichment of the fraction of positive associations in the observed interactome, compared to a randomized interactome, where the differential expression state of the target genes is shuffled. All 54,642 gene-gene interactions where the gene of origin was differentially expressed (p = 0.05) were used. Error bars indicate 95% confidence intervals by bootstrap (10,000 bootstrap). Error bars not overlapping with x-axis indicate p-value < 0.05 by bootstrap.