Search

Biological sequence families contain many sequences that are very similar to each other because they are related by evolution, so the strategy for splitting data into separate training and test sets is a nontrivial choice in benchmarking sequence analysis methods. A random split is insufficient because it will yield test sequences that are closely related or even identical to training sequences. Adapting ideas from independent set graph algorithms, we describe two new methods for splitting sequence data into dissimilar training and test sets. These algorithms input a sequence family and produce a split in which each test sequence is less than p% identical to any individual training sequence. These algorithms successfully split more families than a previous approach, enabling construction of more diverse benchmark datasets.

The genetic code has been proposed to be a “frozen accident”, but the discovery of alternative genetic codes over the past four decades has shown that it can evolve to some degree. Since most examples were found anecdotally, it is difficult to draw general conclusions about the evolutionary trajectories of codon reassignment and why some codons are affected more frequently. To fill in the diversity of genetic codes, we developed Codetta, a computational method to predict the amino acid decoding of each codon from nucleotide sequence data. We surveyed the genetic code usage of over 250,000 bacterial and archaeal genome sequences in GenBank and discovered five new reassignments of arginine codons (AGG, CGA, and CGG), representing the first sense codon changes in bacteria. In a clade of uncultivated Bacilli, the reassignment of AGG to become the dominant methionine codon likely evolved by a change in the amino acid charging of an arginine tRNA. The reassignments of CGA and/or CGG were found in genomes with low GC content, an evolutionary force which likely helped drive these codons to low frequency and enable their reassignment.

SummaryComparisons of genomes of different species are used to identify lineage-specific genes, those genes that appear unique to one species or clade. Lineage-specific genes are often thought to represent genetic novelty that underlies unique adaptations. Identification of these genes depends not only on genome sequences, but also on inferred gene annotations. Comparative analyses typically use available genomes that have been annotated using different methods, increasing the risk that orthologous DNA sequences may be erroneously annotated as a gene in one species but not another, appearing lineage-specific as a result. To evaluate the impact of such “annotation heterogeneity,” we identified four clades of species with sequenced genomes with more than one publicly available gene annotation, allowing us to compare the number of lineage-specific genes inferred when differing annotation methods are used to those resulting when annotation method is uniform across the clade. In these case studies, annotation heterogeneity increases the apparent number of lineage-specific genes by up to 15-fold, suggesting that annotation heterogeneity is a substantial source of potential artifact.

Rfam is a database of RNA families where each of the 3444 families is represented by a multiple sequence alignment of known RNA sequences and a covariance model that can be used to search for additional members of the family. Recent developments have involved expert collaborations to improve the quality and coverage of Rfam data, focusing on microRNAs, viral and bacterial RNAs. We have completed the first phase of synchronising microRNA families in Rfam and miRBase, creating 356 new Rfam families and updating 40. We established a procedure for comprehensive annotation of viral RNA families starting with Flavivirus and Coronaviridae RNAs. We have also increased the coverage of bacterial and metagenome-based RNA families from the ZWD database. These developments have enabled a significant growth of the database, with the addition of 759 new families in Rfam 14. To facilitate further community contribution to Rfam, expert users are now able to build and submit new families using the newly developed Rfam Cloud family curation system. New Rfam website features include a new sequence similarity search powered by RNAcentral, as well as search and visualisation of families with pseudoknots. Rfam is freely available at https://rfam.org.

Genes for which homologs can be detected only in a limited group of evolutionarily related species, called “lineage-specific genes,” are pervasive: Essentially every lineage has them, and they often comprise a sizable fraction of the group’s total genes. Lineage-specific genes are often interpreted as “novel” genes, representing genetic novelty born anew within that lineage. Here, we develop a simple method to test an alternative null hypothesis: that lineage-specific genes do have homologs outside of the lineage that, even while evolving at a constant rate in a novelty-free manner, have merely become undetectable by search algorithms used to infer homology. We show that this null hypothesis is sufficient to explain the lack of detected homologs of a large number of lineage-specific genes in fungi and insects. However, we also find that a minority of lineage-specific genes in both clades are not well explained by this novelty-free model. The method provides a simple way of identifying which lineage-specific genes call for special explanations beyond homology detection failure, highlighting them as interesting candidates for further study., Lineage-specific gene families may arise from evolutionary innovations such as de novo gene origination, or may simply mean that a similarity search program failed to identify more distant homologs. A new computational method for modeling the expected decay of similarity search scores with evolutionary distance allows distinction between the two explanations.

Tavares’ conclusions depend on an assumption that the statistic they use (RAFS) is an appropriate measure of RNA base pair covariation, but RAFS was not designed to measure covariation alone. RAFS detects positive signals in common patterns of primary sequence conservation in absence of any covariation. To illustrate the severity of the problem, we show that Tavares’ analysis reports “significantly covarying base pairs” in 100% identical sequence alignments with no variation or covariation. We use Tavares’ sequence alignment of HOTAIR domain 1 as an example to show that the base pairs they identify as significantly covarying actually arise from primary sequence conservation patterns. Their analysis still reports similar numbers of “significant covarying” base pairs in a negative control in which we permute residues in independent alignment columns to destroy covariation. There remains no significant covariation support for evolutionarily conserved RNA structure in the HOTAIR lncRNA or other lncRNA structures and alignments we have analyzed.

The last few years have witnessed significant changes in Pfam (https://pfam.xfam.org). The number of families has grown substantially to a total of 17,929 in release 32.0. New additions have been coupled with efforts to improve existing families, including refinement of domain boundaries, their classification into Pfam clans, as well as their functional annotation. We recently began to collaborate with the RepeatsDB resource to improve the definition of tandem repeat families within Pfam. We carried out a significant comparison to the structural classification database, namely the Evolutionary Classification of Protein Domains (ECOD) that led to the creation of 825 new families based on their set of uncharacterized families (EUFs). Furthermore, we also connected Pfam entries to the Sequence Ontology (SO) through mapping of the Pfam type definitions to SO terms. Since Pfam has many community contributors, we recently enabled the linking between authorship of all Pfam entries with the corresponding authors’ ORCID identifiers. This effectively permits authors to claim credit for their Pfam curation and link them to their ORCID record.

Pairwise sequence covariations are a signal of conserved RNA secondary structure. We describe a method for distinguishing when lack of covariation signal can be taken as evidence against a conserved RNA structure, as opposed to when a sequence alignment merely has insufficient variation to detect covariations. We find that alignments for several long non-coding RNAs previously shown to lack covariation support do have adequate covariation detection power, providing additional evidence against their proposed conserved structures. Availability and implementation The R-scape web server is at eddylab.org/R-scape, with a link to download the source code. Supplementary information Supplementary data are available at Bioinformatics online.

The HMMER webserver [http://www.ebi.ac.uk/Tools/hmmer] is a free-to-use service which provides fast searches against widely used sequence databases and profile hidden Markov model (HMM) libraries using the HMMER software suite (http://hmmer.org). The results of a sequence search may be summarized in a number of ways, allowing users to view and filter the significant hits by domain architecture or taxonomy. For large scale usage, we provide an application programmatic interface (API) which has been expanded in scope, such that all result presentations are available via both HTML and API. Furthermore, we have refactored our JavaScript visualization library to provide standalone components for different result representations. These consume the aforementioned API and can be integrated into third-party websites. The range of databases that can be searched against has been expanded, adding four sequence datasets (12 in total) and one profile HMM library (6 in total). To help users explore the biological context of their results, and to discover new data resources, search results are now supplemented with cross references to other EMBL-EBI databases.

The MALAT1 (Metastasis-Associated Lung Adenocarcinoma Transcript 1) gene encodes a noncoding RNA that is processed into a long nuclear retained transcript (MALAT1) and a small cytoplasmic tRNA-like transcript (mascRNA). Using an RNA sequence- and structure-based covariance model, we identified more than 130 genomic loci in vertebrate genomes containing the MALAT1 3' end triple-helix structure and its immediate downstream tRNA-like structure, including 44 in the green lizard Anolis carolinensis. Structural and computational analyses revealed a co-occurrence of components of the 3' end module. MALAT1-like genes in Anolis carolinensis are highly expressed in adult testis, thus we named them testis-abundant long noncoding RNAs (tancRNAs). MALAT1-like loci also produce multiple small RNA species, including PIWI-interacting RNAs (piRNAs), from the antisense strand. The 3' ends of tancRNAs serve as potential targets for the PIWI-piRNA complex. Thus, we have identified an evolutionarily conserved class of long noncoding RNAs (lncRNAs) with similar structural constraints, post-transcriptional processing, and subcellular localization and a distinct function in spermatocytes.

The polarized structure of axons and dendrites in neuronal cells depends in part on RNA localization. Previous studies have looked at which polyadenylated RNAs are enriched in neuronal projections or at synapses, but less is known about the distribution of non-adenylated RNAs. By physically dissecting projections from cell bodies of primary rat hippocampal neurons and sequencing total RNA, we found an unexpected set of free circular introns with a non-canonical branchpoint enriched in neuronal projections. These introns appear to be tailless lariats that escape debranching. They lack ribosome occupancy, sequence conservation, and known localization signals, and their function, if any, is not known. Nonetheless, their enrichment in projections has important implications for our understanding of the mechanisms by which RNAs reach distal compartments of asymmetric cells.

The anatomy of many neural circuits is being characterized with increasing resolution, but their molecular properties remain mostly unknown. Here, we characterize gene expression patterns in distinct neural cell types of the Drosophila visual system using genetic lines to access individual cell types, the TAPIN-seq method to measure their transcriptomes, and a probabilistic method to interpret these measurements. We used these tools to build a resource of high-resolution transcriptomes for 100 driver lines covering 67 cell types, available at http://www.opticlobe.com. Combining these transcriptomes with recently reported connectomes helps characterize how information is transmitted and processed across a range of scales, from individual synapses to circuit pathways. We describe examples that include identifying neurotransmitters, including cases of apparent co-release, generating functional hypotheses based on receptor expression, as well as identifying strong commonalities between different cell types., eLife digest In the brain, large numbers of different types of neurons connect with each other to form complex networks. In recent years, researchers have made great progress in mapping all the connections between these cells, creating ‘wiring diagrams’ known as connectomes. However, charting the connections between neurons does not give all the answers as to how the brain works; for example, it does not necessarily reveal the nature of the information two connected cells exchange. Assessing which genes are switched on in different neurons can give insight into neuronal properties that are not obvious from physical connections alone. To fill that knowledge gap, Davis, Nern et al. aimed to measure the genes expressed in a well-characterized network of neurons in the fruit fly visual system. First, 100 fly strains were established, each carrying a single type of neuron colored with a fluorescent marker. Then, a biochemical approach was developed to extract the part of the cell that contains the genetic code from the neurons with the marker. Finally, a statistical tool was used to assess which genes were on in each type of neurons. This led to the creation of a database that shows whether 15,000 genes in each neuron type across 100 fly strains were switched on. Combining this information with previous knowledge about the flies’ visual system revealed new information: for example, it helped to understand which chemicals the neurons use to communicate, and whether certain cells activate or inhibit each other. The work by Davis, Nern et al. demonstrates how genetic approaches can complement other methods, and it offers a new tool for other scientists to use in their work. With more advanced genetic methods, it may one day become possible to better grasp how complex brains in other organisms are organized, and how they are disrupted in disease.

A new method, R-scape, tests whether observed sequence covariation supports a conserved secondary structure in RNA. The program finds no evidence for previously proposed conserved secondary structures in several lncRNAs. Many functional RNAs have an evolutionarily conserved secondary structure. Conservation of RNA base pairing induces pairwise covariations in sequence alignments. We developed a computational method, R-scape (RNA Structural Covariation Above Phylogenetic Expectation), that quantitatively tests whether covariation analysis supports the presence of a conserved RNA secondary structure. R-scape analysis finds no statistically significant support for proposed secondary structures of the long noncoding RNAs HOTAIR, SRA, and Xist.

To infer the sequence and RNA structure specificities of RNA-binding proteins (RBPs) from experiments that enrich for bound sequences, we introduce a convolutional residual network which we call ResidualBind. ResidualBind significantly outperforms previous methods on experimental data from many RBP families. We interrogate ResidualBind to identify what features it has learned from high-affinity sequences with saliency analysis along with 1st-order and 2nd-orderin silicomutagenesis. We show that in addition to sequence motifs, ResidualBind learns a model that includes the number of motifs, their spacing, and both positive and negative effects of RNA structure context. Strikingly, ResidualBind learns RNA structure context, including detailed base-pairing relationships, directly from sequence data, which we confirm on synthetic data. ResidualBind is a powerful, flexible, and interpretable model that can uncovercis-recognition preferences across a broad spectrum of RBPs.

Although convolutional neural networks (CNNs) have been applied to a variety of computational genomics problems, there remains a large gap in our understanding of how they build representations of regulatory genomic sequences. Here we perform systematic experiments on synthetic sequences to reveal how CNN architecture, specifically convolutional filter size and max-pooling, influences the extent that sequence motif representations are learned by first layer filters. We find that CNNs designed to foster hierarchical representation learning of sequence motifs—assembling partial features into whole features in deeper layers—tend to learn distributed representations, i.e. partial motifs. On the other hand, CNNs that are designed to limit the ability to hierarchically build sequence motif representations in deeper layers tend to learn more interpretable localist representations, i.e. whole motifs. We then validate that this representation learning principle established from synthetic sequences generalizes to in vivo sequences., Author summary Although deep convolutional neural networks (CNNs) have demonstrated promise across many regulatory genomics prediction tasks, their inner workings largely remain a mystery. Here we empirically demonstrate how CNN architecture influences the extent that representations of sequence motifs are captured by first layer filters. We find that max-pooling and convolutional filter size modulates information flow, controlling the extent that deeper layers can build features hierarchically. CNNs designed to foster hierarchical representation learning tend to capture partial representations of motifs in first layer filters. On the other hand, CNNs that are designed to limit the ability of deeper layers to hierarchically build upon low-level features tend to learn whole representations of motifs in first layer filters. Together, this study enables the design of CNNs that intentionally learn interpretable representations in easier to access first layer filters (with a small tradeoff in performance), versus building harder to interpret distributed representations, both of which have their strengths and limitations.

Group I catalytic introns have been found in bacterial, viral, organellar, and some eukaryotic genomes, but not in archaea. All known archaeal introns are bulge-helix-bulge (BHB) introns, with the exception of a few group II introns. It has been proposed that BHB introns arose from extinct group I intron ancestors, much like eukaryotic spliceosomal introns are thought to have descended from group II introns. However, group I introns have little sequence conservation, making them difficult to detect with standard sequence similarity searches. Taking advantage of recent improvements in a computational homology search method that accounts for both conserved sequence and RNA secondary structure, we have identified 39 group I introns in a wide range of archaeal phyla, including examples of group I introns and BHB introns in the same host gene.

The HMMER website, available at http://www.ebi.ac.uk/Tools/hmmer/, provides access to the protein homology search algorithms found in the HMMER software suite. Since the first release of the website in 2011, the search repertoire has been expanded to include the iterative search algorithm, jackhmmer. The continued growth of the target sequence databases means that traditional tabular representations of significant sequence hits can be overwhelming to the user. Consequently, additional ways of presenting homology search results have been developed, allowing them to be summarised according to taxonomic distribution or domain architecture. The taxonomy and domain architecture representations can be used in combination to filter the results according to the needs of a user. Searches can also be restricted prior to submission using a new taxonomic filter, which not only ensures that the results are specific to the requested taxonomic group, but also improves search performance. The repertoire of profile hidden Markov model libraries, which are used for annotation of query sequences with protein families and domains, has been expanded to include the libraries from CATH-Gene3D, PIRSF, Superfamily and TIGRFAMs. Finally, we discuss the relocation of the HMMER webserver to the European Bioinformatics Institute and the potential impact that this will have.

It is generally assumed that new genes arise through duplication and/or recombination of existing genes. The probability that a new functional gene could arise out of random non-coding DNA is so far considered to be negligible, since it seems unlikely that such a RNA or protein sequence could have an initial function that influences the fitness of an organism. We have here tested this question systematically, by expressing clones with random sequences in E . coli and subjecting them to competitive growth. Contrary to expectations, we find that random sequences with bioactivity are not rare. In our experiments we find that up to 25% of the evaluated clones enhance the growth rate of their cells and up to 52% inhibit growth. Testing of individual clones in competition assays confirms their activity and provides an indication that their activity could be exerted either by the transcribed RNA or the translated peptide. This suggests that transcribed and translated random parts of the genome could indeed have a high potential to become functional. The results also suggest that random sequences may become an effective new source of molecules for studying cellular functions, as well as for pharmacological activity screening.

The Rfam database (available at http://rfam.xfam.org) is a collection of non-coding RNA families represented by manually curated sequence alignments, consensus secondary structures and annotation gathered from corresponding Wikipedia, taxonomy and ontology resources. In this article, we detail updates and improvements to the Rfam data and website for the Rfam 12.0 release. We describe the upgrade of our search pipeline to use Infernal 1.1 and demonstrate its improved homology detection ability by comparison with the previous version. The new pipeline is easier for users to apply to their own data sets, and we illustrate its ability to annotate RNAs in genomic and metagenomic data sets of various sizes. Rfam has been expanded to include 260 new families, including the well-studied large subunit ribosomal RNA family, and for the first time includes information on short sequence- and structure-based RNA motifs present within families.

Transcriptomics experiments and computational predictions both enable systematic discovery of new functional RNAs. However, many putative noncoding transcripts arise instead from artifacts and biological noise, and current computational prediction methods have high false positive rates. I discuss prospects for improving computational methods for analyzing and identifying functional RNAs, with a focus on detecting signatures of conserved RNA secondary structure. An interesting new front is the application of chemical and enzymatic experiments that probe RNA structure on a transcriptome-wide scale. I review several proposed approaches for incorporating structure probing data into the computational prediction of RNA secondary structure. Using probabilistic inference formalisms, I show how all these approaches can be unified in a well-principled framework, which in turn allows RNA probing data to be easily integrated into a wide range of analyses that depend on RNA secondary structure inference. Such analyses include homology search and genome-wide detection of new structural RNAs.

Rod and cone photoreceptors are highly similar in many respects but they have important functional and molecular differences. Here, we investigate genome-wide patterns of DNA methylation and chromatin accessibility in mouse rods and cones and correlate differences in these features with gene expression, histone marks, transcription factor binding, and DNA sequence motifs. Loss of NR2E3 in rods shifts their epigenomes to a more cone-like state. The data further reveal wide differences in DNA methylation between retinal photoreceptors and brain neurons. Surprisingly, we also find a substantial fraction of DNA hypo-methylated regions in adult rods that are not in active chromatin. Many of these regions exhibit hallmarks of regulatory regions that were active earlier in neuronal development, suggesting that these regions could remain undermethylated due to the highly compact chromatin in mature rods. This work defines the epigenomic landscapes of rods and cones, revealing features relevant to photoreceptor development and function. DOI: http://dx.doi.org/10.7554/eLife.11613.001, eLife digest Vision in humans is made possible by a light-sensing sheet of cells at the back of the eye called the retina. The surface of the retina is populated by specialized sensory cells, known as rods and cones. The rod cells detect very dim light, while the cones are less sensitive to light but are used to detect color. Together, the rods and cones gather the information needed to create a picture that is then transmitted to the brain. Rods and cones have been studied for decades, and genetic analyses have revealed the patterns of gene expression that lead a cell to develop into either a rod or a cone. Researchers have also identified several key regulatory genes that control these patterns, but less is known about the role of other factors that control the expression of genes. Chemical modifications to DNA or modifications to the proteins associated with DNA – which are collectively called epigenetic modifications – can either promote or inhibit the activation of nearby genes. Now, Mo et al. have shown that rods and cones from mice have very different patterns of epigenetic modifications. The experiments also revealed that many sections of DNA that are marked to promote gene activation contain known rod-specific or cone-specific genes; and that rod cells need a known regulatory gene to develop their specific pattern of epigenetic modifications. Finally, Mo et al. showed that epigenetic regulation differed between brain cells and rods and cones. These insights into epigenetic regulation of rod and cone genes may help explain why some people with eye diseases caused by the same genetic mutation may develop symptoms at different ages or lose vision at different rates. The new information about gene regulation may also help scientists to reprogram stem cells to become healthy rods or cones that could be transplanted into people with eye disease to restore their vision. DOI: http://dx.doi.org/10.7554/eLife.11613.002

International audience; In the last two years the Pfam database (http://pfam.xfam.org) has undergone a substantial reorganisation to reduce the effort involved in making a release, thereby permitting more frequent releases. Arguably the most significant of these changes is that Pfam is now primarily based on the UniProtKB reference proteomes, with the counts of matched sequences and species reported on the website restricted to this smaller set. Building families on reference proteomes sequences brings greater stability, which decreases the amount of manual curation required to maintain them. It also reduces the number of sequences displayed on the website, whilst still providing access to many important model organisms. Matches to the full UniProtKB database are, however, still available and Pfam annotations for individual UniProtKB sequences can still be retrieved. Some Pfam entries (1.6%) which have no matches to reference proteomes remain; we are working with UniProt to see if sequences from them can be incorporated into reference proteomes. Pfam-B, the automatically-generated supplement to Pfam, has been removed. The current release (Pfam 29.0) includes 16 295 entries and 559 clans. The facility to view the relationship between families within a clan has been improved by the introduction of a new tool.

We present a database of repetitive DNA elements, called Dfam (http://dfam.janelia.org). Many genomes contain a large fraction of repetitive DNA, much of which is made up of remnants of transposable elements (TEs). Accurate annotation of TEs enables research into their biology and can shed light on the evolutionary processes that shape genomes. Identification and masking of TEs can also greatly simplify many downstream genome annotation and sequence analysis tasks. The commonly used TE annotation tools RepeatMasker and Censor depend on sequence homology search tools such as cross_match and BLAST variants, as well as Repbase, a collection of known TE families each represented by a single consensus sequence. Dfam contains entries corresponding to all Repbase TE entries for which instances have been found in the human genome. Each Dfam entry is represented by a profile hidden Markov model, built from alignments generated using RepeatMasker and Repbase. When used in conjunction with the hidden Markov model search tool nhmmer, Dfam produces a 2.9% increase in coverage over consensus sequence search methods on a large human benchmark, while maintaining low false discovery rates, and coverage of the full human genome is 54.5%. The website provides a collection of tools and data views to support improved TE curation and annotation efforts. Dfam is also available for download in flat file format or in the form of MySQL table dumps.

Many tools are available to analyse genomes but are often challenging to use in a cell type–specific context. We have developed a method similar to the isolation of nuclei tagged in a specific cell type (INTACT) technique [Deal,R.B. and Henikoff,S. (2010) A simple method for gene expression and chromatin profiling of individual cell types within a tissue. Dev. Cell, 18, 1030–1040; Steiner,F.A., Talbert,P.B., Kasinathan,S., Deal,R.B. and Henikoff,S. (2012) Cell-type-specific nuclei purification from whole animals for genome-wide expression and chromatin profiling. Genome Res., doi:10.1101/gr.131748.111], first developed in plants, for use in Drosophila neurons. We profile gene expression and histone modifications in Kenyon cells and octopaminergic neurons in the adult brain. In addition to recovering known gene expression differences, we also observe significant cell type–specific chromatin modifications. In particular, a small subset of differentially expressed genes exhibits a striking anti-correlation between repressive and activating histone modifications. These genes are enriched for transcription factors, recovering those known to regulate mushroom body identity and predicting analogous regulators of octopaminergic neurons. Our results suggest that applying INTACT to specific neuronal populations can illuminate the transcriptional regulatory networks that underlie neuronal cell identity.

The standard approach for single-sequence RNA secondary structure prediction uses a nearest-neighbor thermodynamic model with several thousand experimentally determined energy parameters. An attractive alternative is to use statistical approaches with parameters estimated from growing databases of structural RNAs. Good results have been reported for discriminative statistical methods using complex nearest-neighbor models, including CONTRAfold, Simfold, and ContextFold. Little work has been reported on generative probabilistic models (stochastic context-free grammars [SCFGs]) of comparable complexity, although probabilistic models are generally easier to train and to use. To explore a range of probabilistic models of increasing complexity, and to directly compare probabilistic, thermodynamic, and discriminative approaches, we created TORNADO, a computational tool that can parse a wide spectrum of RNA grammar architectures (including the standard nearest-neighbor model and more) using a generalized super-grammar that can be parameterized with probabilities, energies, or arbitrary scores. By using TORNADO, we find that probabilistic nearest-neighbor models perform comparably to (but not significantly better than) discriminative methods. We find that complex statistical models are prone to overfitting RNA structure and that evaluations should use structurally nonhomologous training and test data sets. Overfitting has affected at least one published method (ContextFold). The most important barrier to improving statistical approaches for RNA secondary structure prediction is the lack of diversity of well-curated single-sequence RNA secondary structures in current RNA databases.