Christopher A. Lavender

RNA-Puzzles: A CASP-like evaluation of RNA three-dimensional structure prediction

We report the results of a first, collective, blind experiment in RNA three-dimensional (3D) structure prediction, encompassing three prediction puzzles. The goals are to assess the leading edge of RNA structure prediction techniques; compare existing methods and tools; and evaluate their relative strengths, weaknesses, and limitations in terms of sequence length and structural complexity. The results should give potential users insight into the suitability of available methods for different applications and facilitate efforts in the RNA structure prediction community in ongoing efforts to improve prediction tools. We also report the creation of an automated evaluation pipeline to facilitate the analysis of future RNA structure prediction exercises.

Opportunities and obstacles for deep learning in biology and medicine

Deep learning describes a class of machine learning algorithms that are capable of combining raw inputs into layers of intermediate features. These algorithms have recently shown impressive results across a variety of domains. Biology and medicine are data-rich disciplines, but the data are complex and often ill-understood. Hence, deep learning techniques may be particularly well suited to solve problems of these fields. We examine applications of deep learning to a variety of biomedical problems—patient classification, fundamental biological processes and treatment of patients—and discuss whether deep learning will be able to transform these tasks or if the biomedical sphere poses unique challenges. Following from an extensive literature review, we find that deep learning has yet to revolutionize biomedicine or definitively resolve any of the most pressing challenges in the field, but promising advances have been made on the prior state of the art. Even though improvements over previous baselines have been modest in general, the recent progress indicates that deep learning methods will provide valuable means for speeding up or aiding human investigation. Though progress has been made linking a specific neural networks prediction to input features, understanding how users should interpret these models to make testable hypotheses about the system under study remains an open challenge. Furthermore, the limited amount of labelled data for training presents problems in some domains, as do legal and privacy constraints on work with sensitive health records. Nonetheless, we foresee deep learning enabling changes at both bench and bedside with the potential to transform several areas of biology and medicine.

Three-dimensional RNA structure refinement by hydroxyl radical probing

Molecular modeling guided by experimentally derived structural information is an attractive approach for three-dimensional structure determination of complex RNAs that are not amenable to study by high-resolution methods. Hydroxyl radical probing (HRP), which is performed routinely in many laboratories, provides a measure of solvent accessibility at individual nucleotides. HRP measurements have, to date, only been used to evaluate RNA models qualitatively. Here we report the development of a quantitative structure refinement approach using HRP measurements to drive discrete molecular dynamics simulations for RNAs ranging in size from 80 to 230 nucleotides. We first used HRP reactivities to identify RNAs that form extensive helical packing interactions. For these RNAs, we achieved highly significant structure predictions given the inputs of RNA sequence and base pairing. This HRP-directed tertiary structure refinement approach generates robust structural hypotheses that are useful for guiding explorations of structure-function inter-relationships in RNA.

Single-molecule correlated chemical probing of RNA

Significance RNA molecules function as the central conduit of information transfer in biology. To do this, they encode information both in their sequences and in their higher-order structures. Understanding the higher-order structure of RNA remains challenging. In this work we devise a simple, experimentally concise, and accurate approach for examining higher-order RNA structure by converting widely used massively parallel sequencing into an easily implemented single-molecule experiment for detecting through-space interactions and multiple conformations. We then use this experiment to analyze higher-order RNA structure, detect biologically important hidden states, and refine accurate three-dimensional structure models. Complex higher-order RNA structures play critical roles in all facets of gene expression; however, the through-space interaction networks that define tertiary structures and govern sampling of multiple conformations are poorly understood. Here we describe single-molecule RNA structure analysis in which multiple sites of chemical modification are identified in single RNA strands by massively parallel sequencing and then analyzed for correlated and clustered interactions. The strategy thus identifies RNA interaction groups by mutational profiling (RING-MaP) and makes possible two expansive applications. First, we identify through-space interactions, create 3D models for RNAs spanning 80–265 nucleotides, and characterize broad classes of intramolecular interactions that stabilize RNA. Second, we distinguish distinct conformations in solution ensembles and reveal previously undetected hidden states and large-scale structural reconfigurations that occur in unfolded RNAs relative to native states. RING-MaP single-molecule nucleic acid structure interrogation enables concise and facile analysis of the global architectures and multiple conformations that govern function in RNA.

In-cell SHAPE reveals that free 30S ribosome subunits are in the inactive state

Significance It has been known for decades that purified small subunits of the ribosome can interconvert between active and inactive conformations in experiments performed under simplified conditions, but the physiological relevance of this switch has remained unclear. We probed the structure of ribosomal RNA in healthy living cells and discovered that stably assembled 30S subunits exist predominantly in the inactive conformation, with structural differences localized in the functionally important decoding region. Disrupting the ability to interconvert between active and inactive conformations compromised translation in cells. In-cell RNA structure probing supports a model in which “inactive” 30S subunits comprise an abundant in-cell state that regulates ribosome function. It was shown decades ago that purified 30S ribosome subunits readily interconvert between “active” and “inactive” conformations in a switch that involves changes in the functionally important neck and decoding regions. However, the physiological significance of this conformational change had remained unknown. In exponentially growing Escherichia coli cells, RNA SHAPE probing revealed that 16S rRNA largely adopts the inactive conformation in stably assembled, mature 30S subunits and the active conformation in translating (70S) ribosomes. Inactive 30S subunits bind mRNA as efficiently as active subunits but initiate translation more slowly. Mutations that inhibited interconversion between states compromised translation in vivo. Binding by the small antibiotic paromomycin induced the inactive-to-active conversion, consistent with a low-energy barrier between the two states. Despite the small energetic barrier between states, but consistent with slow translation initiation and a functional role in vivo, interconversion involved large-scale changes in structure in the neck region that likely propagate across the 30S body via helix 44. These findings suggest the inactive state is a biologically relevant alternate conformation that regulates ribosome function as a conformational switch.

Widespread transcriptional pausing and elongation control at enhancers

Regulation by gene-distal enhancers is critical for cell type-specific and condition-specific patterns of gene expression. Thus, to understand the basis of gene activity in a given cell type or tissue, we must identify the precise locations of enhancers and functionally characterize their behaviors. Here, we demonstrate that transcription is a nearly universal feature of enhancers in Drosophila and mammalian cells and that nascent RNA sequencing strategies are optimal for identification of both enhancers and superenhancers. We dissect the mechanisms governing enhancer transcription and discover remarkable similarities to transcription at protein-coding genes. We show that RNA polymerase II (RNAPII) undergoes regulated pausing and release at enhancers. However, as compared with mRNA genes, RNAPII at enhancers is less stable and more prone to early termination. Furthermore, we found that the level of histone H3 Lys4 (H3K4) methylation at enhancers corresponds to transcriptional activity such that highly active enhancers display H3K4 trimethylation rather than the H3K4 monomethylation considered a hallmark of enhancers. Finally, our work provides insights into the unique characteristics of superenhancers, which stimulate high-level gene expression through rapid pause release; interestingly, this property renders associated genes resistant to the loss of factors that stabilize paused RNAPII.

Structure-Based Alignment and Consensus Secondary Structures for Three HIV-Related RNA Genomes.

HIV and related primate lentiviruses possess single-stranded RNA genomes. Multiple regions of these genomes participate in critical steps in the viral replication cycle, and the functions of many RNA elements are dependent on the formation of defined structures. The structures of these elements are still not fully understood, and additional functional elements likely exist that have not been identified. In this work, we compared three full-length HIV-related viral genomes: HIV-1NL4-3, SIVcpz, and SIVmac (the latter two strains are progenitors for all HIV-1 and HIV-2 strains, respectively). Model-free RNA structure comparisons were performed using whole-genome structure information experimentally derived from nucleotide-resolution SHAPE reactivities. Consensus secondary structures were constructed for strongly correlated regions by taking into account both SHAPE probing structural data and nucleotide covariation information from structure-based alignments. In these consensus models, all known functional RNA elements were recapitulated with high accuracy. In addition, we identified multiple previously unannotated structural elements in the HIV-1 genome likely to function in translation, splicing and other replication cycle processes; these are compelling targets for future functional analyses. The structure-informed alignment strategy developed here will be broadly useful for efficient RNA motif discovery.

Model-Free RNA Sequence and Structure Alignment Informed by SHAPE Probing Reveals a Conserved Alternate Secondary Structure for 16S rRNA

Discovery and characterization of functional RNA structures remains challenging due to deficiencies in de novo secondary structure modeling. Here we describe a dynamic programming approach for model-free sequence comparison that incorporates high-throughput chemical probing data. Based on SHAPE probing data alone, ribosomal RNAs (rRNAs) from three diverse organisms – the eubacteria E. coli and C. difficile and the archeon H. volcanii – could be aligned with accuracies comparable to alignments based on actual sequence identity. When both base sequence identity and chemical probing reactivities were considered together, accuracies improved further. Derived sequence alignments and chemical probing data from protein-free RNAs were then used as pseudo-free energy constraints to model consensus secondary structures for the 16S and 23S rRNAs. There are critical differences between these experimentally-informed models and currently accepted models, including in the functionally important neck and decoding regions of the 16S rRNA. We infer that the 16S rRNA has evolved to undergo large-scale changes in base pairing as part of ribosome function. As high-quality RNA probing data become widely available, structurally-informed sequence alignment will become broadly useful for de novo motif and function discovery.

Downstream Antisense Transcription Predicts Genomic Features That Define the Specific Chromatin Environment at Mammalian Promoters.

Antisense transcription is a prevalent feature at mammalian promoters. Previous studies have primarily focused on antisense transcription initiating upstream of genes. Here, we characterize promoter-proximal antisense transcription downstream of gene transcription starts sites in human breast cancer cells, investigating the genomic context of downstream antisense transcription. We find extensive correlations between antisense transcription and features associated with the chromatin environment at gene promoters. Antisense transcription downstream of promoters is widespread, with antisense transcription initiation observed within 2 kb of 28% of gene transcription start sites. Antisense transcription initiates between nucleosomes regularly positioned downstream of these promoters. The nucleosomes between gene and downstream antisense transcription start sites carry histone modifications associated with active promoters, such as H3K4me3 and H3K27ac. This region is bound by chromatin remodeling and histone modifying complexes including SWI/SNF subunits and HDACs, suggesting that antisense transcription or resulting RNA transcripts contribute to the creation and maintenance of a promoter-associated chromatin environment. Downstream antisense transcription overlays additional regulatory features, such as transcription factor binding, DNA accessibility, and the downstream edge of promoter-associated CpG islands. These features suggest an important role for antisense transcription in the regulation of gene expression and the maintenance of a promoter-associated chromatin environment.

ORIO (Online Resource for Integrative Omics): a web-based platform for rapid integration of next generation sequencing data

Abstract Established and emerging next generation sequencing (NGS)-based technologies allow for genome-wide interrogation of diverse biological processes. However, accessibility of NGS data remains a problem, and few user-friendly resources exist for integrative analysis of NGS data from different sources and experimental techniques. Here, we present Online Resource for Integrative Omics (ORIO; https://orio.niehs.nih.gov/), a web-based resource with an intuitive user interface for rapid analysis and integration of NGS data. To use ORIO, the user specifies NGS data of interest along with a list of genomic coordinates. Genomic coordinates may be biologically relevant features from a variety of sources, such as ChIP-seq peaks for a given protein or transcription start sites from known gene models. ORIO first iteratively finds read coverage values at each genomic feature for each NGS dataset. Data are then integrated using clustering-based approaches, giving hierarchical relationships across NGS datasets and separating individual genomic features into groups. In focusing its analysis on read coverage, ORIO makes limited assumptions about the analyzed data; this allows the tool to be applied across data from a variety of experiments and techniques. Results from analysis are presented in dynamic displays alongside user-controlled statistical tests, supporting rapid statistical validation of observed results. We emphasize the versatility of ORIO through diverse examples, ranging from NGS data quality control to characterization of enhancer regions and integration of gene expression information. Easily accessible on a public web server, we anticipate wide use of ORIO in genome-wide investigations by life scientists.