Visualization of RNA structure models within the Integrative Genomics Viewer

Abstract

Analyses of the interrelationships between RNA structure and function are increasingly important components of genomic studies. The SHAPE-MaP strategy enables accurate RNA structure probing and realistic structure modeling of kilobase-length noncoding RNAs and mRNAs. Existing tools for visualizing RNA structure models are not suitable for efficient analysis of long, structurally heterogeneous RNAs. In addition, structure models are often advantageously interpreted in the context of other experimental data and gene annotation information, for which few tools currently exist. We have developed a module within the widely used and well supported open-source Integrative Genomics Viewer (IGV) that allows visualization of SHAPE and other chemical probing data, including raw reactivities, data-driven structural entropies, and data-constrained base-pair secondary structure models, in context with linear genomic data tracks. We illustrate the usefulness of visualizing RNA structure in the IGV by exploring structure models for a large viral RNA genome, comparing bacterial mRNA structure in cells with its structure under cell- and protein-free conditions, and comparing a noncoding RNA structure modeled using SHAPE data with a base-pairing model inferred through sequence covariation analysis.

Keywords: IGV; RNA structure; SHAPE-MaP; entropy; pairing probabilities.

FIGURE 1.

IGV screen images illustrating exploration of RNA structure in an HIV-1 genome. (A) Overview of the entire ∼9200-nt genome. (B) Zoomed-in view of the highly structured ∼350-nt RRE and nearby regions. (C) View of a relatively unstructured region in the sequence encoding the env protein. Base-pair arcs are colored by estimated pairing probability (green, blue, and yellow: >80%, 30%, and 10%, respectively). Secondary structures and base pairing probabilities were generated with SuperFold (Smola et al. 2015). The maximum base-pairing distance was set to 600 nt; windowed SHAPE reactivities and (Shannon) entropies were computed as 55-nt windowed medians; and windowed SHAPE reactivities are plotted centered about the global median.

FIGURE 2.

E. coli mRNA structure visualized under two experimental conditions: in-cell versus protein- and cell-free, using the 1M7 reagent that yields robust structure signals under in-cell and cell-free conditions (Tyrrell et al. 2013, McGinnis et al. 2015, Smola et al. 2015, 2016). IGV screen images are shown. Base pairs, SHAPE reactivities, and entropies are shown as in Figure 1.

FIGURE 3.

Conserved structures in the untranslated region of the E. coli rpmH gene. (A) View spanning the 5′-UTR and first coding region. In this case, conserved structures (purple bar) are in a region predicted to be highly structured (low SHAPE reactivity) and well defined (low Shannon entropy). (B) Fully zoomed-in view of conserved structures in the untranslated region upstream of the rpmH gene. There is strong agreement between the measured SHAPE reactivity profile and the derived secondary structure model. Base pairs and SHAPE reactivities are shown as in Figure 1. Images shown are zoomed-in views based on Figure 2.

FIGURE 4.

Comparison between IGV tracks illustrating a secondary structure model for the E. coli 6S RNA based on experimental SHAPE data versus a structure model from the Rfam database containing base pairs inferred using nucleotide conservation and covariation (Gardner et al. 2009). Note that, although the SHAPE-directed model includes a greater number of base pairs than included in the Rfam model, all base pairs in the SHAPE-informed model are consistent with the nucleotide-resolution chemical probing data.