Assessing Outlier Probabilities in Transcriptomics Data When Evaluating a Classifier.

Outliers in the training or test set used to fit and evaluate a classifier on transcriptomics data can considerably change the estimated performance of the model. Hence, an either too weak or a too optimistic accuracy is then reported and the estimated model performance cannot be reproduced on independent data. It is then also doubtful whether a classifier qualifies for clinical usage. We estimate classifier performances in simulated gene expression data with artificial outliers and in two real-world datasets. As a new approach, we use two outlier detection methods within a bootstrap procedure to estimate the outlier probability for each sample and evaluate classifiers before and after outlier removal by means of cross-validation. We found that the removal of outliers changed the classification performance notably. For the most part, removing outliers improved the classification results. Taking into account the fact that there are various, sometimes unclear reasons for a sample to be an outlier, we strongly advocate to always report the performance of a transcriptomics classifier with and without outliers in training and test data. This provides a more diverse picture of a classifier's performance and prevents reporting models that later turn out to be not applicable for clinical diagnoses.

Deciphering Eurasian otter (Lutra lutra L.) and seal (Phoca vitulina L.; Halichoerus grypus F.) diet: Metabarcoding tailored for fresh and saltwater fish species.

Long-lived top predators shape biodiversity structure in their ecosystems and predator-prey interactions are critical in decoding how communities function. Studies on the foraging ecology of seals and Eurasian otters in Western Europe are outdated and most studies solely performed traditional hard part analysis. Molecular metabarcoding can be used as an innovative noninvasive diet analysis tool, which has proven efficient and complementary to hard part analysis, however, lacking application in the wider North Sea area. In this study, DNA from digesta, collected between 2014-2020, were used to identify fish species in the diet of 47 Eurasian otters, 54 harbour seals and 21 grey seals by applying a next-generation metabarcoding approach. A newly designed 16S rRNA primer, providing the best coverage of >130 local marine and freshwater fish species, was used to amplify prey DNA from seal scats and otter gut content sampled from the North Sea and regional freshwater bodies. Frequent fish species included tench, ninespine stickleback and white bream in otters; hooknose and common roach in grey seals and Pleuronectidae and sand gobies in harbour seals. Bipartite network analysis showed a strong overlap of harbour and grey seal diets. Otter diet intersected with both seal species in terms of freshwater species. This study provides new knowledge about dietary composition and community assemblage of fish prey in otters and seals in the North Sea and regional freshwaters, and a new molecular tool to elucidate predator-prey interactions and interspecies competition in complex and changing ecosystems under pressure from anthropogenic activities.

LUCS: a high-resolution nucleic acid sequencing tool for accurate long-read analysis of individual DNA molecules.

Nucleic acid sequence analyses are fundamental to all aspects of biological research, spanning aging, mitochondrial DNA (mtDNA) and cancer, as well as microbial and viral evolution. Over the past several years, significant improvements in DNA sequencing, including consensus sequence analysis, have proven invaluable for high-throughput studies. However, all current DNA sequencing platforms have limited utility for studies of complex mixtures or of individual long molecules, the latter of which is crucial to understanding evolution and consequences of single nucleotide variants and their combinations. Here we report a new technology termed LUCS (Long-molecule UMI-driven Consensus Sequencing), in which reads from third-generation sequencing are aggregated by unique molecular identifiers (UMIs) specific for each individual DNA molecule. This enables in-silico reconstruction of highly accurate consensus reads of each DNA molecule independent of other molecules in the sample. Additionally, use of two UMIs enables detection of artificial recombinants (chimeras). As proof of concept, we show that application of LUCS to assessment of mitochondrial genomes in complex mixtures from single cells was associated with an error rate of 1X10-4 errors/nucleotide. Thus, LUCS represents a major step forward in DNA sequencing that offers high-throughput capacity and high-accuracy reads in studies of long DNA templates and nucleotide variants in heterogenous samples.