Deep learning models for RNA secondary structure prediction (probably) do not generalize across families.

MOTIVATION: The secondary structure of RNA is of importance to its function. Over the last few years, several papers attempted to use machine learning to improve de novo RNA secondary structure prediction. Many of these papers report impressive results for intra-family predictions but seldom address the much more difficult (and practical) inter-family problem. RESULTS: We demonstrate that it is nearly trivial with convolutional neural networks to generate pseudo-free energy changes, modelled after structure mapping data that improve the accuracy of structure prediction for intra-family cases. We propose a more rigorous method for inter-family cross-validation that can be used to assess the performance of learning-based models. Using this method, we further demonstrate that intra-family performance is insufficient proof of generalization despite the widespread assumption in the literature and provide strong evidence that many existing learning-based models have not generalized inter-family. AVAILABILITY AND IMPLEMENTATION: Source code and data are available at https://github.com/marcellszi/dl-rna. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Nutrient stress can have opposite effects on the ability of plants to tolerate foliar herbivory and floral herbivory.

Discovering how organisms respond to the combinations of stressors they face in their environment is an enduring challenge for ecologists. A particular focus has been how natural enemies and abiotic stressors faced by plants may interact in their effect on the ecology and evolution of plant defense strategies. Here, we report on the results of an experiment measuring how reproduction in the clonal herbaceous plant horsenettle (Solanum carolinense) is affected by damage by leaf-feeding and by flower-feeding herbivores-as well as how horsenettle's tolerance of these different types of herbivory may be altered by nutrient stress. Leaf herbivory by lace bugs reduced horsenettle's seed production and root growth, and the relative impacts were greater in fertilized than in nutrient-stressed plants. In contrast, simulated-floral herbivory reduced seed production to a similar degree in fertilized and nutrient-stressed plants. However, compensation for floral herbivory through increased root growth occurred to a much greater extent in the fertilized than in the nutrient-stressed plants. These results can be explained in terms of the limiting resource model of plant tolerance, with leaf damage interpreted as exacerbating carbon limitation in the fertilized plants and floral damage ameliorating carbon limitation in the fertilized plants. These results can be extended to predicting patterns in the field: Although plants in a nutrient-poor environment may have overall low fitness, they are likely to be more tolerant of leaf herbivores-though this benefit may be countered by lower tolerance of any floral herbivores that share the environment.

From rags to enriched: metagenomic insights into ammonia-oxidizing archaea following ammonia enrichment of a denuded oligotrophic soil ecosystem.

Stored topsoil acts as a microbial inoculant for ecological restoration of land after disturbance, but the altered circumstances frequently create unfavourable conditions for microbial survival. Nitrogen cycling is a critical indicator for ecological success and this study aimed to investigate the cornerstone taxa driving the process. Previous in silico studies investigating stored topsoil discovered persistent archaeal taxa with the potential for re-establishing ecological activity. Ammonia oxidization is the limiting step in nitrification and as such, ammonia-oxidizing archaea (AOA) can be considered one of the gatekeepers for the re-establishment of the nitrogen cycle in disturbed soils. Semi-arid soil samples were enriched with ammonium sulfate to promote the selective enrichment of ammonia oxidizers for targeted genomic recovery, and to investigate the microbial response of the microcosm to nitrogen input. Ammonia addition produced an increase in AOA population, particularly within the genus Candidatus Nitrosotalea, from which metagenome-assembled genomes (MAGs) were successfully recovered. The Ca. Nitrosotalea archaeon candidates' ability to survive in extreme conditions and rapidly respond to ammonia input makes it a potential bioprospecting target for application in ecological restoration of semi-arid soils and the recovered MAGs provide a metabolic blueprint for developing potential strategies towards isolation of these acclimated candidates.

Small investments with big returns: environmental genomic bioprospecting of microbial life.

Microorganisms and their natural products are major drivers of ecological processes and industrial applications. Microbial bioprospecting has been critical for the advancement in various fields such as pharmaceuticals, sustainable industries, food security and bioremediation. Next generation sequencing has been paramount in the exploration of diverse environmental microbiomes. It presents a culture-independent approach to investigating hitherto uncultured taxa, resulting in the creation of massive sequence databases, which are available in the public domain. Genome mining searches available (meta)genomic data for target biosynthetic genes, and combined with the large-scale public data, this in-silico bioprospecting method presents an efficient and extensive way to uncover microbial bioproducts. Bioinformatic tools have progressed to a stage where we can recover genomes from the environment; these metagenome-assembled genomes present a way to understand the metabolic capacity of microorganisms in a physiological and ecological context. Environmental sampling been extensive across various ecological settings, including microbiomes with unique physicochemical properties that could influence the discovery of novel functions and metabolic pathways. Although in-silico methods cannot completely substitute in-vitro studies, the contextual information it provides is invaluable for understanding the ecological and taxonomic distribution of microbial genotypes and to form effective strategies for future microbial bioprospecting efforts.

Cold, clumpy accretion onto an active supermassive black hole.

Supermassive black holes in galaxy centres can grow by the accretion of gas, liberating energy that might regulate star formation on galaxy-wide scales. The nature of the gaseous fuel reservoirs that power black hole growth is nevertheless largely unconstrained by observations, and is instead routinely simplified as a smooth, spherical inflow of very hot gas. Recent theory and simulations instead predict that accretion can be dominated by a stochastic, clumpy distribution of very cold molecular clouds--a departure from the 'hot mode' accretion model--although unambiguous observational support for this prediction remains elusive. Here we report observations that reveal a cold, clumpy accretion flow towards a supermassive black hole fuel reservoir in the nucleus of the Abell 2597 Brightest Cluster Galaxy (BCG), a nearby (redshift z = 0.0821) giant elliptical galaxy surrounded by a dense halo of hot plasma. Under the right conditions, thermal instabilities produce a rain of cold clouds that fall towards the galaxy's centre, sustaining star formation amid a kiloparsec-scale molecular nebula that is found at its core. The observations show that these cold clouds also fuel black hole accretion, revealing 'shadows' cast by the molecular clouds as they move inward at about 300 kilometres per second towards the active supermassive black hole, which serves as a bright backlight. Corroborating evidence from prior observations of warmer atomic gas at extremely high spatial resolution, along with simple arguments based on geometry and probability, indicate that these clouds are within the innermost hundred parsecs of the black hole, and falling closer towards it.

Mutations of Helicobacter pylori RdxA are mainly related to the phylogenetic origin of the strain and not to metronidazole resistance.

OBJECTIVES: Drug resistance of Helicobacter pylori is a major clinical problem worldwide. The objective of the present study was to investigate the prevalence of antibiotic-resistant H. pylori in the city of Shenzhen in China, as well as to identify the genetic mutations specifically associated with drug resistance rather than unrelated phylogenetic signals. METHODS: Antibiotic susceptibility testing was performed on 238 clinical strains successfully isolated from H. pylori-positive dyspeptic patients who underwent gastroscopy at the Department of Gastroenterology in Shenzhen People's Second Hospital. Following WGS of all strains using Illumina technology, mutation and phylogenetic analyses were performed. RESULTS: The resistance rates were 84.9%, 35.3%, 25.2% and 2.1% for metronidazole, clarithromycin, ciprofloxacin and rifampicin, respectively. An A2143G conversion in the 23S rRNA gene was the primary mutation observed in clarithromycin-resistant strains, whilst N87K/I and D91G/N/Y in GyrA were detected in ciprofloxacin-resistant strains. In RdxA, our results demonstrated that only R16H/C and M21A are significant contributors to metronidazole resistance; there were 15 other sites, but these are phylogenetically related and thus unrelated to metronidazole resistance. CONCLUSIONS: There is a high prevalence of metronidazole, clarithromycin and ciprofloxacin resistance and a low prevalence of rifampicin resistance in H. pylori from Shenzhen, China. Omission of phylogenetically related sites will help to improve identification of sites genuinely related to antibiotic resistance in H. pylori and, we believe, other species.

Determining parameters for non-linear models of multi-loop free energy change.

MOTIVATION: Predicting the secondary structure of RNA is a fundamental task in bioinformatics. Algorithms that predict secondary structure given only the primary sequence, and a model to evaluate the quality of a structure, are an integral part of this. These algorithms have been updated as our model of RNA thermodynamics changed and expanded. An exception to this has been the treatment of multi-loops. Although more advanced models of multi-loop free energy change have been suggested, a simple, linear model has been used since the 1980s. However, recently, new dynamic programing algorithms for secondary structure prediction that could incorporate these models were presented. Unfortunately, these models appear to have lower accuracy for secondary structure prediction. RESULTS: We apply linear regression and a new parameter optimization algorithm to find better parameters for the existing linear model and advanced non-linear multi-loop models. These include the Jacobson-Stockmayer and Aalberts & Nandagopal models. We find that the current linear model parameters may be near optimal for the linear model, and that no advanced model performs better than the existing linear model parameters even after parameter optimization. AVAILABILITY AND IMPLEMENTATION: Source code and data is available at https://github.com/maxhwardg/advanced_multiloops. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Recent progress in the structure of glycogen serving as a durable energy reserve in bacteria.

Glycogen is conventionally considered as a transient energy reserve that can be rapidly synthesized for glucose accumulation and mobilized for ATP production. However, this conception is not completely applicable to prokaryotes due to glycogen structural heterogeneity. A number of studies noticed that glycogen with small average chain length gc in bacteria has the potential to degrade slowly, which might prolong bacterial environment survival. This phenomenon was previously examined and later formulated as the durable energy storage mechanism hypothesis. Although recent research has been warming to the hypothesis, experimental validation is still missing at current stage. In this review, we summarized recent progress of the hypothesis, provided a supporting mathematical model, and explored the technical pitfalls that shall be avoided in glycogen study.

Molecular Structure of Glycogen in Escherichia coli.

Glycogen, a randomly branched glucose polymer, provides energy storage in organisms. It forms small ß particles which in animals bind to form composite α particles, which give better glucose release. Simulations imply ß particle size is controlled only by activities and sizes of glycogen biosynthetic enzymes and sizes of polymer chains. Thus, storing more glucose requires forming more ß particles, which are expected to sometimes form α particles. No α particles have been reported in bacteria, but the extraction techniques might have caused degradation. Using milder glycogen extraction techniques on Escherichia coli, transmission electron microscopy and size-exclusion chromatography showed α particles, consistent with this hypothesis for α-particle formation. Molecular density and size distributions show similarities with animal glycogen, despite very different metabolic processes. These general polymer constraints are such that any organism which needs to store and then release glucose will have similar α and ß particle structures: a type of convergent evolution.

Advanced multi-loop algorithms for RNA secondary structure prediction reveal that the simplest model is best.

Algorithmic prediction of RNA secondary structure has been an area of active inquiry since the 1970s. Despite many innovations since then, our best techniques are not yet perfect. The workhorses of the RNA secondary structure prediction engine are recursions first described by Zuker and Stiegler in 1981. These have well understood caveats; a notable flaw is the ad-hoc treatment of multi-loops, also called helical-junctions, that persists today. While several advanced models for multi-loops have been proposed, it seems to have been assumed that incorporating them into the recursions would lead to intractability, and so no algorithms for these models exist. Some of these models include the classical model based on Jacobson-Stockmayer polymer theory, and another by Aalberts and Nadagopal that incorporates two-length-scale polymer physics. We have realized practical, tractable algorithms for each of these models. However, after implementing these algorithms, we found that no advanced model was better than the original, ad-hoc model used for multi-loops. While this is unexpected, it supports the praxis of the current model.

Genomic epidemiology and population structure of Neisseria gonorrhoeae from remote highly endemic Western Australian populations.

BACKGROUND: Neisseria gonorrhoeae causes gonorrhoea, the second most commonly notified sexually transmitted infection in Australia. One of the highest notification rates of gonorrhoea is found in the remote regions of Western Australia (WA). Unlike isolates from the major Australian population centres, the remote community isolates have low rates of antimicrobial resistance (AMR). Population structure and whole-genome comparison of 59 isolates from the Western Australian N. gonorrhoeae collection were used to investigate relatedness of isolates cultured in the metropolitan and remote areas. Core genome phylogeny, multilocus sequencing typing (MLST), N. gonorrhoeae multi-antigen sequence typing (NG-MAST) and N. gonorrhoeae sequence typing for antimicrobial resistance (NG-STAR) in addition to hierarchical clustering of sequences were used to characterize the isolates. RESULTS: Population structure analysis of the 59 isolates together with 72 isolates from an international collection, revealed six population groups suggesting that N. gonorrhoeae is a weakly clonal species. Two distinct population groups, Aus1 and Aus2, represented 63% of WA isolates and were mostly composed of the remote community isolates that carried no chromosomal AMR genotypes. In contrast, the Western Australian metropolitan isolates were frequently multi-drug resistant and belonged to population groups found in the international database, suggesting international transmission of the isolates. CONCLUSIONS: Our study suggests that the population structure of N. gonorrhoeae is distinct between the communities in remote and metropolitan WA. Given the high rate of AMR in metropolitan regions, ongoing surveillance is essential to ensure the enduring efficacy of the empiric gonorrhoea treatment in remote WA.

Defense with benefits? Ducking plants outperformed erect plants in the goldenrod Solidago gigantea in the absence of herbivory.

PREMISE OF THE STUDY: Despite the fact that herbivores can be highly detrimental to their host plants' fitness, plant populations often maintain genetic variation for resistance to their natural enemies. Investigating the various costs (e.g., allocation tradeoffs, autotoxicity, and ecological costs) that may prevent plants from evolving to their fullest potential resistance has been a productive strategy for shedding insight into the eco-evolutionary dynamics of plant-herbivore communities. METHODS: Recent studies have shown that some individuals of goldenrod (Solidago spp.) evade apex-attacking herbivores by a temporary nodding of their stem (i.e., resistance-by-ducking). Although ducking provides an obvious fitness benefit to these individuals, nonducking (erect) morphs persist in goldenrod populations. In this study, I investigated potential costs of ducking in Solidago gigantea in terms of tradeoffs involving growth and reproduction in a common garden experiment using field-collected seeds. KEY RESULTS: The S. gigantea population contained substantial genetic variation for stem morph, with 28% erect and 72% ducking stems. In the absence of herbivory, ducking plants were taller, had thicker stems, and produced an average of 20% more seeds than erect plants. CONCLUSIONS: This study suggests that resistance-by-ducking, instead of being costly, actually comes with additional, nondefense-related benefits. These results support the conclusion that the factors that constrain the evolution of resistance in plant populations are likely to be more subtle and complex than simple tradeoffs in resource allocation.

Constraints on the evolution of resistance to gall flies in Solidago altissima: resistance sometimes costs more than it is worth.

Plant populations frequently maintain submaximal levels of resistance to natural enemies, even in the presence of substantial genetic variation for resistance. Identifying constraints on the evolution of increased resistance has been a major goal of researchers of plant-herbivore interactions. In a glasshouse study, we measured relative costs and benefits of resistance of tall goldenrod (Solidago altissima) to the gall-inducing tephritid Eurosta solidaginis. We exposed multiple ramets of 26 goldenrod genets to nutrient or shade stress and to oviposition by E. solidaginis. The presence of a gall cost a ramet an average of 1743 seeds, but the cost differed 10-fold across environments. Plant genets varied widely for an induced 'hypersensitive' response in which meristem cells become necrotic and kill E. solidaginis hatchlings before gall induction. There was no evidence that this highly effective resistance trait carried an allocation cost. However, the response carried a risk of autotoxicity, as necrosis killed the apex of 37% of the ungalled ramets. On average, a damaged apex cost each ramet 5015 seeds. Autotoxicity may constrain the resistance of S. altissima to an intermediate level, and variation in environmental conditions may alter the relative costs and benefits of resistance and tolerance, thus maintaining genetic variation within goldenrod populations.

Novel Moraxella catarrhalis prophages display hyperconserved non-structural genes despite their genomic diversity.

BACKGROUND: Moraxella catarrhalis is an important pathogen that often causes otitis media in children, a disease that is not currently vaccine preventable. Asymptomatic colonisation of the human upper respiratory tract is common and lack of clearance by the immune system is likely due to the emergence of seroresistant genetic lineages. No active bacteriophages or prophages have been described in this species. This study was undertaken to identify and categorise prophages in M. catarrhalis, their genetic diversity and the relationship of such diversity with the host-species phylogeny. RESULTS: This study presents a comparative analysis of 32 putative prophages identified in 95 phylogenetically variable, newly sequenced M. catarrhalis genomes. The prophages were genotypically classified into four diverse clades. The genetic synteny of each clade is similar to the group 1 phage family Siphoviridae, however, they form genotypic clusters that are distinct from other members of this family. No core genetic sequences exist across the 32 prophages despite clades 2, 3, and 4 sharing the most sequence identity. The analysis of non-structural prophage genes (coding the integrase, and terminase), and portal gene showed that the respective genes were identical for clades 2, 3, and 4, but unique for clade 1. Empirical analysis calculated that these genes are unexpectedly hyperconserved, under purifying selection, suggesting a tightly regulated functional role. As such, it is improbable that the prophages are decaying remnants but stable components of a fluctuating, flexible and unpredictable system ultimately maintained by functional constraints on non-structural and packaging genes. Additionally, the plate encoding genes were well conserved across all four prophage clades, and the tail fibre genes, commonly responsible for receptor recognition, were clustered into three major groups distributed across the prophage clades. A pan-genome of 283,622 bp was identified, and the prophages were mapped onto the diverse M. catarrhalis multi-locus sequence type (MLST) backbone. CONCLUSION: This study has provided the first evidence of putatively mobile prophages in M. catarrhalis, identifying a diverse and fluctuating system dependent on the hyperconservation of a few key, non-structural genes. Some prophages harbour virulence-related genes, and potentially influence the physiology and virulence of M. catarrhalis. Importantly our data will provide supporting information on the identification of novel prophages in other species by adding greater weight to the identification of non-structural genes.

Influence of in situ progressive N-terminal is still controversial truncation of glycogen branching enzyme in Escherichia coli DH5α on glycogen structure, accumulation, and bacterial viability.

BACKGROUND: Glycogen average chain length (ACL) has been linked with bacterial durability, but this was on the basis of observations across different species. We therefore wished to investigate the relationship between bacterial durability and glycogen ACL by varying glycogen average chain length in a single species. It has been shown that progressive shortening of the N-terminus of glycogen branching enzyme (GBE) leads to a lengthening of oligosaccharide inter-α-1,6-glycosidic chain lengths, so we sought to harness this to create a set of Escherichia coli DH5α strains with a range of glycogen average chain lengths, and assess these strains for durability related attributes, such as starvation, cold and desiccation stress resistance, and biofilm formation. RESULTS: A series of Escherichia coli DH5α mutants were created with glgB genes that were in situ progressively N-terminus truncated. N-terminal truncation shifted the distribution of glycogen chain lengths from 5-11 DP toward 13-50 DP, but the relationship between glgB length and glycogen ACL was not linear. Surprisingly, removal of the first 270 nucleotides of glgB (glgBΔ270) resulted in comparatively high glycogen accumulation, with the glycogen having short ACL. Complete knockout of glgB led to the formation of amylose-like glycogen containing long, linear α1,4-glucan chains with significantly reduced branching frequency. Physiologically, the set of mutant strains had reduced bacterial starvation resistance, while minimally increasing bacterial desiccation resistance. Finally, although there were no obvious changes in cold stress resistance or biofilm forming ability, one strain (glgBΔ180) had significantly increased biofilm formation in favourable media. CONCLUSIONS: Despite glgB being the first gene of an operon, it is clear that in situ mutation is a viable means to create more biologically relevant mutant strains. Secondly, there was the suggestion in the data that impairments of starvation, cold and desiccation resistance were worse for the strain lacking glgB, though the first of these was not statistically significant. The results provide prima facie evidence linking abiotic stress tolerance with shorter glycogen ACL. However, further work needs to be done, perhaps in a less labile species. Further work is also required to tease out the complex relationship between glycogen abundance and glycogen structure.

Consistent gene expression profiles in MexTAg transgenic mouse and wild type mouse asbestos-induced mesothelioma.

BACKGROUND: The MexTAg transgenic mouse model of mesothelioma replicates many aspects of human mesothelioma, including induction by asbestos, pathogenicity and response to cytotoxic chemotherapy, despite high levels of the SV40 large T Antigen (TAg) in the mesothelial compartment. This model enables analysis of the molecular events associated with asbestos induced mesothelioma and is utilised here to investigate the molecular dynamics of tumours induced in these mice, using gene expression patterns as a read out. METHODS: Gene expression of MexTAg mesothelioma cell lines bearing a high or low number of copies of the TAg transgene were compared to wild type mouse mesotheliomas and normal mouse mesothelial cells using Affymetrix microarray. These data were then compared to a similar published human microarray study using the same platform. RESULTS: The main expression differences between transgenic mouse and wild type mouse mesotheliomas occurred for genes involved in cell cycle regulation and DNA replication, as would be expected from overexpression of the TAg oncogene. Quantitative PCR confirmed that E2F and E2F regulated genes were significantly more upregulated in MexTAg mesotheliomas and MexTAg mesothelial cells compared to wild type mesotheliomas. Like human mesothelioma, both MexTAg and wild type mesotheliomas had more genes underexpressed than overexpressed compared to normal mouse mesothelial cells. Most notably, the cdkn2 locus was deleted in the wild type mouse mesotheliomas, consistent with 80 % human mesotheliomas, however, this region was not deleted in MexTAg mesotheliomas. Regardless of the presence of TAg, all mouse mesotheliomas had a highly concordant set of deregulated genes compared to normal mesothelial cells that overlapped with the deregulated genes between human mesotheliomas and mesothelial cells. CONCLUSIONS: This investigation demonstrates that the MexTAg mesotheliomas are comparable with wild type mouse mesotheliomas in their representation of human mesothelioma at the molecular level, with some key gene expression differences that are attributable to the TAg transgene expression. Of particular note, MexTAg mesothelioma development was not dependent on cdkn2 deletion.

Comparative analysis of the full genome of Helicobacter pylori isolate Sahul64 identifies genes of high divergence.

Isolates of Helicobacter pylori can be classified phylogeographically. High genetic diversity and rapid microevolution are a hallmark of H. pylori genomes, a phenomenon that is proposed to play a functional role in persistence and colonization of diverse human populations. To provide further genomic evidence in the lineage of H. pylori and to further characterize diverse strains of this pathogen in different human populations, we report the finished genome sequence of Sahul64, an H. pylori strain isolated from an indigenous Australian. Our analysis identified genes that were highly divergent compared to the 38 publically available genomes, which include genes involved in the biosynthesis and modification of lipopolysaccharide, putative prophage genes, restriction modification components, and hypothetical genes. Furthermore, the virulence-associated vacA locus is a pseudogene and the cag pathogenicity island (cagPAI) is not present. However, the genome does contain a gene cluster associated with pathogenicity, including dupA. Our analysis found that with the addition of Sahul64 to the 38 genomes, the core genome content of H. pylori is reduced by approximately 14% (∼170 genes) and the pan-genome has expanded from 2,070 to 2,238 genes. We have identified three putative horizontally acquired regions, including one that is likely to have been acquired from the closely related Helicobacter cetorum prior to speciation. Our results suggest that Sahul64, with the absence of cagPAI, highly divergent cell envelope proteins, and a predicted nontransportable VacA protein, could be more highly adapted to ancient indigenous Australian people but with lower virulence potential compared to other sequenced and cagPAI-positive H. pylori strains.

Heuristic RNA pseudoknot prediction including intramolecular kissing hairpins.

Pseudoknots are an essential feature of RNA tertiary structures. Simple H-type pseudoknots have been studied extensively in terms of biological functions, computational prediction, and energy models. Intramolecular kissing hairpins are a more complex and biologically important type of pseudoknot in which two hairpin loops form base pairs. They are hard to predict using free energy minimization due to high computational requirements. Heuristic methods that allow arbitrary pseudoknots strongly depend on the quality of energy parameters, which are not yet available for complex pseudoknots. We present an extension of the heuristic pseudoknot prediction algorithm DotKnot, which covers H-type pseudoknots and intramolecular kissing hairpins. Our framework allows for easy integration of advanced H-type pseudoknot energy models. For a test set of RNA sequences containing kissing hairpins and other types of pseudoknot structures, DotKnot outperforms competing methods from the literature. DotKnot is available as a web server under http://dotknot.csse.uwa.edu.au.

Predicting pseudoknotted structures across two RNA sequences.

MOTIVATION: Laboratory RNA structure determination is demanding and costly and thus, computational structure prediction is an important task. Single sequence methods for RNA secondary structure prediction are limited by the accuracy of the underlying folding model, if a structure is supported by a family of evolutionarily related sequences, one can be more confident that the prediction is accurate. RNA pseudoknots are functional elements, which have highly conserved structures. However, few comparative structure prediction methods can handle pseudoknots due to the computational complexity. RESULTS: A comparative pseudoknot prediction method called DotKnot-PW is introduced based on structural comparison of secondary structure elements and H-type pseudoknot candidates. DotKnot-PW outperforms other methods from the literature on a hand-curated test set of RNA structures with experimental support. AVAILABILITY: DotKnot-PW and the RNA structure test set are available at the web site http://dotknot.csse.uwa.edu.au/pw. CONTACT: janaspe@csse.uwa.edu.au SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.