Binary classifiers are routinely evaluated with performance measures such as sensitivity and specificity, and performance is frequently illustrated with Receiver Operating Characteristics (ROC) plots. Alternative measures such as positive predictive value (PPV) and the associated Precision/Recall (PRC) plots are used less frequently. Many bioinformatics studies develop and evaluate classifiers that are to be applied to strongly imbalanced datasets in which the number of negatives outweighs the number of positives significantly. While ROC plots are visually appealing and provide an overview of a classifier's performance across a wide range of specificities, one can ask whether ROC plots could be misleading when applied in imbalanced classification scenarios. We show here that the visual interpretability of ROC plots in the context of imbalanced datasets can be deceptive with respect to conclusions about the reliability of classification performance, owing to an intuitive but wrong interpretation of specificity. PRC plots, on the other hand, can provide the viewer with an accurate prediction of future classification performance due to the fact that they evaluate the fraction of true positives among positive predictions. Our findings have potential implications for the interpretation of a large number of studies that use ROC plots on imbalanced datasets.

SummaryThe precision–recall plot is more informative than the ROC plot when evaluating classifiers on imbalanced datasets, but fast and accurate curve calculation tools for precision–recall plots are currently not available. We have developed Precrec, an R library that aims to overcome this limitation of the plot. Our tool provides fast and accurate precision–recall calculations together with multiple functionalities that work efficiently under different conditions.Availability and ImplementationPrecrec is licensed under GPL-3 and freely available from CRAN (https://cran.r-project.org/package=precrec). It is implemented in R with C ++.Supplementary information Supplementary data are available at Bioinformatics online.

MicroRNAs (miRNAs) are a class of small noncoding RNAs that can regulate many genes by base pairing to sites in mRNAs. The functionality of miRNAs overlaps that of short interfering RNAs (siRNAs), and many features of miRNA targeting have been revealed experimentally by studying miRNA-mimicking siRNAs. This review outlines the features associated with animal miRNA targeting and describes currently available prediction tools.

MicroRNAs (miRNAs) regulate genes post transcription by pairing with messenger RNA (mRNA). Variants such as single nucleotide polymorphisms (SNPs) in miRNA regulatory regions might result in altered protein levels and disease. Genome-wide association studies (GWAS) aim at identifying genomic regions that contain variants associated with disease, but lack tools for finding causative variants. We present a computational tool that can help identifying SNPs associated with diseases, by focusing on SNPs affecting miRNA-regulation of genes. The tool predicts the effects of SNPs in miRNA target sites and uses linkage disequilibrium to map these miRNA-related variants to SNPs of interest in GWAS. We compared our predicted SNP effects in miRNA target sites with measured SNP effects from allelic imbalance sequencing. Our predictions fit measured effects better than effects based on differences in free energy or differences of TargetScan context scores. We also used our tool to analyse data from published breast cancer and Parkinson's disease GWAS and significant trait-associated SNPs from the NHGRI GWAS Catalog. A database of predicted SNP effects is available at http://www.bigr.medisin.ntnu.no/mirsnpscore/. The database is based on haplotype data from the CEU HapMap population and miRNAs from miRBase 16.0.

Feeding plant‐based diet through smoltification of Atlantic salmon requires verification of the optimal level of 1C nutrients. Here, we fed Atlantic salmon plant‐based diets containing three different surplus amounts of the 1C nutrients; methionine, cobalamin (vitamin B12), pyridoxine (vitamin B6) and folic acid during 6 weeks in fresh water, through smoltification, followed by 3 months on‐growing period in salt water. The three diets were fed to fish dispersed in triplicate tanks throughout the experiment. Mean start body weight was 32 g. Dietary methionine levels in the diets were 6.7, 9.2 and 11.7 g/kg. Dietary B6 was 6.75, 8.45 and 11 mg/kg. Cobalamin was 0.16, 0.18 and 0.20 mg/kg. While dietary folic acid was 2.9, 4.8 and 6.3 mg/kg, diets are referred to as low, medium and high 1C diet. All other amino acids were similar between diets. The results showed no differences in growth or feed utilization in the fresh water period, but following the on‐growing salt water period, differences between diets occurred. The fish fed the medium 1C diet showed better growth, as compared to fish fed the low or high 1C diet (p = .009). The medium 1C fed fish showed a relative lower liver weight compared with fish fed low or high 1C diet (p = .025). Condition factor was better in fish fed the medium and high 1C diet as compared to those fed the low 1C diet (p = .0006). As expected, free methionine in liver, plasma and muscle increased by dietary methionine inclusion. Surplus vitamins only had minor effect on tissue concentrations. Based on these findings, we conclude that the micronutrient and methionine level presented in the medium 1C diet improved the growth, liver size and condition factor; however, more research is needed to evaluate the optimal requirement level for each of the 1C nutrients.

Micronutrients (vitamins and minerals) have been less well studied compared to macronutrients (fats, proteins, and carbohydrates) although they play important roles in growth, metabolism, and maintenance of tissues. Hence, there is growing interest to understand the influence of micronutrients across various aspects in nutritional research. In the last two decades, aquaculture feeds have been shifted to containing more plant-based materials to meet the increasing demand and maintain the sustainability in the industry. A recent whole life cycle feeding trial of Atlantic salmon ( Salmo salar ) with graded levels of micronutrient packages has concluded that the levels of several B-vitamins and microminerals need to be increased from the current recommendation levels for optimal growth and fish welfare when plant-based diets are used. Here, we show the effect of micronutrient supplementation on hepatic transcriptional and epigenetic regulation in a dose dependent manner. . Specifically, our aim is to reveal the mechanisms of altered cell metabolism, which results in improved growth performance by micronutrient surpluses, at gene expression and DNA methylation levels. Our results strongly indicate that micronutrient supplementation suppresses gene expression in lipid metabolism in a dose-dependent manner and broadly affects DNA methylation in cell-adhesion and cell-signalling. In particular, it increases DNA methylation levels on the acetyl-CoA carboxylase alpha promoter in a concentration-dependent manner, which further suggests that acetyl-CoA carboxylase alpha is an upstream epigenetic regulator controlling its downstream lipid biosynthesis activities. This study demonstrates a comprehensive analysis to reveal an important role of micronutrients in lipid metabolism through epigenetic control of gene expression.

MicroRNAs (miRNAs) play a key role in regulating mRNA expression, and individual miRNAs have been proposed as diagnostic and therapeutic candidates. The identification of such candidates is complicated by the involvement of multiple miRNAs and mRNAs as well as unknown disease topology of the miRNAs. Here, we investigated if disease-associated miRNAs regulate modules of disease-associated mRNAs, if those miRNAs act complementarily or synergistically, and if single or combinations of miRNAs can be targeted to alter module functions. We first analyzed publicly available miRNA and mRNA expression data for five different diseases. Integrated target prediction and network-based analysis showed that the miRNAs regulated modules of disease-relevant genes. Most of the miRNAs acted complementarily to regulate multiple mRNAs. To functionally test these findings, we repeated the analysis using our own miRNA and mRNA expression data from CD4+ T cells from patients with seasonal allergic rhinitis. This is a good model of complex diseases because of its well-defined phenotype and pathogenesis. Combined computational and functional studies confirmed that miRNAs mainly acted complementarily and that a combination of two complementary miRNAs, miR-223 and miR-139-3p, could be targeted to alter disease-relevant module functions, namely, the release of type 2 helper T-cell (Th2) cytokines. Taken together, our findings indicate that miRNAs act complementarily to regulate modules of disease-related mRNAs and can be targeted to alter disease-relevant functions.

BackgroundMicroRNA (miRNA) target genes tend to have relatively long and conserved 3' untranslated regions (UTRs), but to what degree these characteristics contribute to miRNA targeting is poorly understood. Different high-throughput experiments have, for example, shown that miRNAs preferentially regulate genes with both short and long 3' UTRs and that target site conservation is both important and irrelevant for miRNA targeting.ResultsWe have analyzed several gene context-dependent features, including 3' UTR length, 3' UTR conservation, and messenger RNA (mRNA) expression levels, reported to have conflicting influence on miRNA regulation. By taking into account confounding factors such as technology-dependent experimental bias and competition between transfected and endogenous miRNAs, we show that two factors - target gene expression and competition - could explain most of the previously reported experimental differences. Moreover, we find that these and other target site-independent features explain about the same amount of variation in target gene expression as the target site-dependent features included in the TargetScan model.ConclusionsOur results show that it is important to consider confounding factors when interpreting miRNA high throughput experiments and urge special caution when using microarray data to compare average regulatory effects between groups of genes that have different average gene expression levels.