We describe an update of MirGeneDB, the manually curated microRNA gene database. Adhering to uniform and consistent criteria for microRNA annotation and nomenclature, we substantially expanded MirGeneDB with 30 additional species representing previously missing metazoan phyla such as sponges, jellyfish, rotifers and flatworms. MirGeneDB 2.1 now consists of 75 species spanning over ∼800 million years of animal evolution, and contains a total number of 16 670 microRNAs from 1549 families. Over 6000 microRNAs were added in this update using ∼550 datasets with ∼7.5 billion sequencing reads. By adding new phylogenetically important species, especially those relevant for the study of whole genome duplication events, and through updating evolutionary nodes of origin for many families and genes, we were able to substantially refine our nomenclature system. All changes are traceable in the specifically developed MirGeneDB version tracker. The performance of read-pages is improved and microRNA expression matrices for all tissues and species are now also downloadable. Altogether, this update represents a significant step toward a complete sampling of all major metazoan phyla, and a widely needed foundation for comparative microRNA genomics and transcriptomics studies. MirGeneDB 2.1 is part of RNAcentral and Elixir Norway, publicly and freely available at http://www.mirgenedb.org/.

MicroRNAs (miRNAs) are small non-coding RNAs that regulate gene expression by binding to partially complementary regions within the 3’UTR of their target genes. Computational methods play an important role in target prediction and assume that the miRNA “seed region” (nt 2 to 8) is required for functional targeting, but typically only identify ∼80% of known bindings. Recent studies have highlighted a role for the entire miRNA, suggesting that a more flexible methodology is needed. We present a novel approach for miRNA target prediction based on Deep Learning (DL) which, rather than incorporating any knowledge (such as seed regions), investigates the entire miRNA and 3’TR mRNA nucleotides to learn a uninhibited set of feature descriptors related to the targeting process. We collected more than 150,000 experimentally validated homo sapiens miRNA:gene targets and cross referenced them with different CLIP-Seq, CLASH and iPAR-CLIP datasets to obtain ∼20,000 validated miRNA:gene exact target sites. Using this data, we implemented and trained a deep neural network—composed of autoencoders and a feed-forward network—able to automatically learn features describing miRNA-mRNA interactions and assess functionality. Predictions were then refined using information such as site location or site accessibility energy. In a comparison using independent datasets, our DL approach consistently outperformed existing prediction methods, recognizing the seed region as a common feature in the targeting process, but also identifying the role of pairings outside this region. Thermodynamic analysis also suggests that site accessibility plays a role in targeting but that it cannot be used as a sole indicator for functionality. Data and source code available at: https://bitbucket.org/account/user/bipous/projects/MIRAW. microRNAs are small RNA molecules that regulate biological processes by binding to the 3’UTR of a gene and their dysregulation is associated with several diseases. Computationally predicting these targets remains a challenge as they only partially match their target and so there can be hundreds of targets for a single microRNA. Current tools assume that most of the knowledge defining a microRNA-gene interaction can be captured by analysing the binding produced in the seed region (∼ the first 8nt in the miRNA). However, recent studies show that the whole microRNA can be important and form non-canonical targets. Here, we use a target prediction methodology that relies on deep neural networks to automatically learn the relevant features describing microRNA-gene interactions for predicting microRNA targets. This means we make no assumptions about what is important, leaving the task to the deep neural network. A key part of the work is obtaining a suitable dataset. Thus, we collected and curated more than 150,000 experimentally verified microRNA targets and used them to train the network. Using this approach, we are able to gain a better understanding of non-canonical targets and to improve the accuracy of state-of-the-art prediction tools.

Japanese encephalitis (JE) is an arthropod-borne disease associated with the majority of viral encephalitis cases in the Asia-Pacific region. The causative agent, Japanese encephalitis virus (JEV), has been phylogenetically divided into five genotypes. Recent surveillance data indicate that genotype I (GI) is gradually replacing genotype III (GIII) as the dominant genotype. To investigate the mechanism behind the genotype shift and the potential consequences in terms of vaccine efficacy, human cases, and virus dissemination, we collected (i) all full-length and partial JEV molecular sequences and (ii) associated genotype and host information comprising a data set of 873 sequences. We then examined differences between the two genotypes at the genetic and epidemiological level by investigating amino acid mutations, positive selection, and host range.IMPORTANCEWe found that although GI is dominant, it has fewer sites predicted to be under positive selection, a narrower host range, and significantly fewer human isolates. For the E protein, the sites under positive selection define a haplotype set for each genotype that shows striking differences in their composition and diversity, with GIII showing significantly more variety than GI. Our results suggest that GI has displaced GIII by achieving a replication cycle that is more efficient but is also more restricted in its host range.Japanese encephalitis is an arthropod-borne disease associated with the majority of viral encephalitis cases in the Asia-Pacific region. The causative agent, Japanese encephalitis virus (JEV), has been divided into five genotypes based on sequence similarity. Recent data indicate that genotype I (GI) is gradually replacing genotype III (GIII) as the dominant genotype. Understanding the reasons behind this shift and the potential consequences in terms of vaccine efficacy, human cases, and virus dissemination is important for controlling the spread of the virus and reducing human fatalities. We collected all available full-length and partial JEV molecular sequences and associated genotype and host information. We then examined differences between the two genotypes at the genetic and epidemiological levels by investigating amino acid mutations, positive selection, and host range. Our results suggest that GI has displaced GIII by achieving a replication cycle that is more efficient but more restricted in host range.

Over the last few years, the number of microRNAs in the human genome has become a controversially debated issue. Several publications reported thousands of putative novel microRNAs not included in the curated microRNA gene database MirGeneDB and the repository miRBase. Recently, by using sequencing of ∼300 human tissues and cell lines, the human RNA atlas, an expanded inventory of human RNA annotations, was published, reporting thousands of putative microRNAs. We, the developers of established microRNA prediction tools and hosts of MirGeneDB, raise concerns about the frequently applied prediction and functional validation strategies, briefly discussing the drawbacks of false positive detections. By means of quantifying well-established biogenesis-derived features, we show that the reported novel microRNAs essentially represent false-positives and argue that the human microRNA complement, at about 550 microRNA genes, is already near complete. Output of available tools must be curated as false predictions will misguide scientists looking for biomarkers or therapeutic targets.

High-throughput sequencing is increasingly favoured to assay the presence and abundance of microRNAs (miRNAs) in biological samples, even from low RNA amounts, and a number of commercial vendors now offer kits that allow miRNA sequencing from sub-nanogram (ng) inputs. Although biases introduced during library preparation have been documented, the relative performance of current reagent kits has not been investigated in detail. Here, six commercial kits capable of handling <100ng total RNA input were used for library preparation, performed by kit manufactures, on synthetic miRNAs of known quantities and human total RNA samples. We compared the performance of miRNA detection sensitivity, reliability, titration response and the ability to detect differentially expressed miRNAs. In addition, we assessed the use of unique molecular identifiers (UMI) sequence tags in one kit. We observed differences in detection sensitivity and ability to identify differentially expressed miRNAs between the kits, but none were able to detect the full repertoire of synthetic miRNAs. The reliability within the replicates of all kits was good, while larger differences were observed between the kits, although none could accurately quantify the relative levels of the majority of miRNAs. UMI tags, at least within the input ranges tested, offered little advantage to improve data utility. In conclusion, biases in miRNA abundance are heavily influenced by the kit used for library preparation, suggesting that comparisons of datasets prepared by different procedures should be made with caution. This article is intended to assist researchers select the most appropriate kit for their experimental conditions.