Exploration and analysis of R-loop mapping data with RLBase

Abstract

R-loops are three-stranded nucleic acid structures formed from the hybridization of RNA and DNA. In 2012, Ginno et al. introduced the first R-loop mapping method. Since that time, dozens of R-loop mapping studies have been conducted, yielding hundreds of publicly available datasets. Current R-loop databases provide only limited access to these data. Moreover, no web tools for analyzing user-supplied R-loop datasets have yet been described. In our recent work, we reprocessed 810 R-loop mapping samples, building the largest R-loop data resource to date. We also defined R-loop consensus regions and developed a framework for R-loop data analysis. Now, we introduce RLBase, a user-friendly database that provides the capability to (i) explore hundreds of public R-loop mapping datasets, (ii) explore R-loop consensus regions, (iii) analyze user-supplied data and (iv) download standardized and reprocessed datasets. RLBase is directly accessible via the following URL: https://gccri.bishop-lab.uthscsa.edu/shiny/rlbase/.

Figure 1.

RLBase overview. (A) Graphical illustration showing the processing steps for RLBase. 810 R-loop datasets were downloaded and reprocessed using the RLPipes, RLHub and RLSeq software. RLBase is an R-shiny web database that was developed based on these software and data. (B) Graphical illustration depicting the core functionality of RLBase. RLBase provides the capability to (i) explore R-loop mapping samples and generate summary visualizations, (ii) explore R-loop regions and view them in the genome browser, (iii) analyze user-supplied R-loop mapping data in the browser and (iv) download standardized and reprocessed R-loop mapping data.

Figure 2.

Select visualizations generated from RLBase. (A–E) Visualizations produced by RLBase with all modes, ‘Show labeled controls’ and ‘Show predicted control’ and ‘hg38’ genome options selected in the ‘Table controls’. Additionally, the DRIPc sample SRX1070676 was selected in the ‘RLBase Samples Table’. (A) Donut charts showing the proportions of samples by Mode, Label and Prediction. (B) The sample-sample correlation heatmap from the ‘Sample-sample comparison’ panel. (C) Genomic feature plots showing the distribution of ENCODE cis-regulatory element (CRE) feature enrichment within ‘POS’ (predicted to map R-loops) and ‘NEG’ (not predicted to map R-loops) samples. The user-selected sample (SRX1070676) is highlighted in each feature. Abbreviations: CTCF, CTCF binding site; enhD, distal enhancer; enhP, enhancer–promoter; K4m3, H3K4me3 histone modification site; prom, promoter site. (D) RLFS analysis plot with P value from permutation testing. (E) Venn diagram showing the overlap between selected sample ranges (SRX1070676 peaks) and RL regions. P value and odds ratio from Fisher’s exact test.

Figure 3.

RLBase genome browser session. The browser session includes representative RLBase sample coverage tracks alongside S9.6 and dRNH consensus signal, RL regions, CpG islands, R-loop forming sequences (RLFS) and ENCODE CREs. The screen capture shown here was taken in the area surrounding the CALM2 gene.