Dataset Information

Biases in small RNA deep sequencing data.

ABSTRACT: High-throughput RNA sequencing (RNA-seq) is considered a powerful tool for novel gene discovery and fine-tuned transcriptional profiling. The digital nature of RNA-seq is also believed to simplify meta-analysis and to reduce background noise associated with hybridization-based approaches. The development of multiplex sequencing enables efficient and economic parallel analysis of gene expression. In addition, RNA-seq is of particular value when low RNA expression or modest changes between samples are monitored. However, recent data uncovered severe bias in the sequencing of small non-protein coding RNA (small RNA-seq or sRNA-seq), such that the expression levels of some RNAs appeared to be artificially enhanced and others diminished or even undetectable. The use of different adapters and barcodes during ligation as well as complex RNA structures and modifications drastically influence cDNA synthesis efficacies and exemplify sources of bias in deep sequencing. In addition, variable specific RNA G/C-content is associated with unequal polymerase chain reaction amplification efficiencies. Given the central importance of RNA-seq to molecular biology and personalized medicine, we review recent findings that challenge small non-protein coding RNA-seq data and suggest approaches and precautions to overcome or minimize bias.

SUBMITTER: Raabe CA

PROVIDER: S-EPMC3919602 | biostudies-literature | 2014 Feb

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Biases in small RNA deep sequencing data.

Raabe Carsten A CA Tang Thean-Hock TH Brosius Juergen J Rozhdestvensky Timofey S TS

Nucleic acids research 20131105 3

High-throughput RNA sequencing (RNA-seq) is considered a powerful tool for novel gene discovery and fine-tuned transcriptional profiling. The digital nature of RNA-seq is also believed to simplify meta-analysis and to reduce background noise associated with hybridization-based approaches. The development of multiplex sequencing enables efficient and economic parallel analysis of gene expression. In addition, RNA-seq is of particular value when low RNA expression or modest changes between samples ...[more]

PMID: 24198247

Similar Datasets

Project description:Deep generative models, such as variational autoencoders (VAEs) or deep Boltzmann machines (DBMs), can generate an arbitrary number of synthetic observations after being trained on an initial set of samples. This has mainly been investigated for imaging data but could also be useful for single-cell transcriptomics (scRNA-seq). A small pilot study could be used for planning a full-scale experiment by investigating planned analysis strategies on synthetic data with different sample sizes. It is unclear whether synthetic observations generated based on a small scRNA-seq dataset reflect the properties relevant for subsequent data analysis steps. We specifically investigated two deep generative modeling approaches, VAEs and DBMs. First, we considered single-cell variational inference (scVI) in two variants, generating samples from the posterior distribution, the standard approach, or the prior distribution. Second, we propose single-cell deep Boltzmann machines (scDBMs). When considering the similarity of clustering results on synthetic data to ground-truth clustering, we find that the [Formula: see text] variant resulted in high variability, most likely due to amplifying artifacts of small datasets. All approaches showed mixed results for cell types with different abundance by overrepresenting highly abundant cell types and missing less abundant cell types. With increasing pilot dataset sizes, the proportions of the cells in each cluster became more similar to that of ground-truth data. We also showed that all approaches learn the univariate distribution of most genes, but problems occurred with bimodality. Across all analyses, in comparing 10[Formula: see text] Genomics and Smart-seq2 technologies, we could show that for 10[Formula: see text] datasets, which have higher sparsity, it is more challenging to make inference from small to larger datasets. Overall, the results show that generative deep learning approaches might be valuable for supporting the design of scRNA-seq experiments.

Project description:BackgroundAlthough advances in sequencing technologies have popularized the use of microRNA (miRNA) sequencing (miRNA-seq) for the quantification of miRNA expression, questions remain concerning the optimal methodologies for analysis and utilization of the data. The construction of a miRNA sequencing library selects RNA by length rather than type. However, as we have previously described, miRNAs represent only a subset of the species obtained by size selection. Consequently, the libraries obtained for miRNA sequencing also contain a variety of additional species of small RNAs. This study looks at the prevalence of these other species obtained from bone marrow aspirate specimens and explores the predictive value of these small RNAs in the determination of response to therapy in myelodysplastic syndromes (MDS).MethodsPaired pre and post treatment bone marrow aspirate specimens were obtained from patients with MDS who were treated with either azacytidine or decitabine (24 pre-treatment specimens, 23 post-treatment specimens) with 22 additional non-MDS control specimens. Total RNA was extracted from these specimens and submitted for next generation sequencing after an additional size exclusion step to enrich for small RNAs. The species of small RNAs were enumerated, single nucleotide variants (SNVs) identified, and finally the differential expression of tRNA-derived species (tDRs) in the specimens correlated with diseasestatus and response to therapy.ResultsUsing miRNA sequencing data generated from bone marrow aspirate samples of patients with known MDS (N = 47) and controls (N = 23), we demonstrated that transfer RNA (tRNA) fragments (specifically tRNA halves, tRHs) are one of the most common species of small RNA isolated from size selection. Using tRNA expression values extracted from miRNA sequencing data, we identified six tRNA fragments that are differentially expressed between MDS and normal samples. Using the elastic net method, we identified four tRNAs-derived small RNAs (tDRs) that together can explain 67 % of the variation in treatment response for MDS patients. Similar analysis of specifically mitochondrial tDRs (mt-tDRs) identified 13 mt-tDRs which distinguished disease status in the samples and a single mt-tDR which predited response. Finally, 14 SNVs within the tDRs were found in at least 20 % of the MDS samples and were not observed in any of the control specimens.DiscussionThis study highlights the prevalence of tDRs in RNA-seq studies focused on small RNAs. The potential etiologies of these species, both technical and biologic, are discussed as well as important challenges in the interpretation of tDR data.ConclusionsOur analysis results suggest that tRNA fragments can be accurately detected through miRNA sequencing data and that the expression of these species may be useful in the diagnosis of MDS and the prediction of response to therapy.

Dataset Information

Biases in small RNA deep sequencing data.

Publications

Biases in small RNA deep sequencing data.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets