Controlling False Positive Rates in Methods for Differential Gene Expression Analysis using RNA-Seq Data

David M Rocke , Luyao Ruan , J. Jared Gossett , Blythe Durbin-Johnson , Sharon Aviran
doi: http://dx.doi.org/10.1101/018739

We review existing methods for the analysis of RNA-Seq data and place them in a common framework of a sequence of tasks that are usually part of the process. We show that many existing methods produce large numbers of false positives in cases where the null hypothesis is true by construction and where actual data from RNA-Seq studies are used, as opposed to simulations that make specific assumptions about the nature of the data. We show that some of those mathematical assumptions about the data likely are one of the causes of the false positives, and define a general structure that is not apparently subject to these problems. The best performance was shown by limma-voom and by some simple methods composed of easily understandable steps.

Fine-mapping cellular QTLs with RASQUAL and ATAC-seq

Natsuhiko Kumasaka , Andrew Knights , Daniel Gaffney
doi: http://dx.doi.org/10.1101/018788

When cellular traits are measured using high-throughput DNA sequencing quantitative trait loci (QTLs) manifest at two levels: population level differences between individuals and allelic differences between cis-haplotypes within individuals. We present RASQUAL (Robust Allele Specific QUAntitation and quality controL), a novel statistical approach for association mapping that integrates genetic effects and robust modelling of biases in next generation sequencing (NGS) data within a single, probabilistic framework. RASQUAL substantially improves causal variant localisation and sensitivity of association detection over existing methods in RNA-seq, DNaseI-seq and ChIP-seq data. We illustrate how RASQUAL can be used to maximise association detection by generating the first map of chromatin accessibility QTLs (caQTLs) in a European population using ATAC-seq. Despite a modest sample size, we identified 2,706 independent caQTLs (FDR 10%) and illustrate how RASQUAL’s improved causal variant localisation provides powerful information for fine-mapping disease-associated variants. We also map “multipeak” caQTLs, identical genetic associations found across multiple, independent open chromatin regions and illustrate how genetic signals in ATAC-seq data can be used to link distal regulatory elements with gene promoters. Our results highlight how joint modelling of population and allele-specific genetic signals can improve functional interpretation of noncoding variation.

Surveying the relative impact of mRNA features on local ribosome profiling read density in 28 datasets.

Patrick O’Connor , Dmitry Andreev , Pavel Baranov
doi: http://dx.doi.org/10.1101/018762

Ribosome profiling is a promising technology for exploring gene expression. However, ribosome profiling data are characterized by a substantial number of outliers due to technical and biological factors. Here we introduce a simple computational method, Ribo-seq Unit Step Transformation (RUST) for the characterization of ribosome profiling data. We show that RUST is robust and outperforms conventional normalization techniques in the presence of sporadic noise. We used RUST to analyse 28 publicly available ribosome profiling datasets obtained from mammalian cells and tissues and from yeast. This revealed substantial protocol dependent variation in the composition of footprint libraries. We selected a high quality dataset to explore the mRNA features that affect local decoding rates and found that the amino acid identity encoded by the codon in the A-site is the major contributing factor followed by the identity of the codon itself and then the amino acid in the P-site. We also found that bulky amino acids slow down ribosome movement when they occur within the peptide tunnel and Proline residues may decrease or increase ribosome velocities depending on the context in which they occur. Moreover we show that a few parameters obtained with RUST are sufficient for predicting experimental densities with high accuracy. Due to its robustness and low computational demand, RUST could be used for quick routine characterization of ribosome profiling datasets to assess their quality as well as for the analysis of the relative impact of mRNA sequence features on local decoding rates.

Distinct nucleosome distribution patterns in two structurally and functionally differentiated nuclei of a unicellular eukaryote

Jie Xiong , Shan Gao , Wen Dui , Wentao Yang , Xiao Chen , Sean D Taverna , Ronald E. Pearlman , Wendy Ashlock , Wei Miao , Yifan Liu
doi: http://dx.doi.org/10.1101/018754

The ciliate protozoan Tetrahymena thermophila contains two types of structurally and functionally differentiated nuclei: the transcriptionally active somatic macronucleus (MAC) and the transcriptionally silent germ-line micronucleus (MIC). Here we demonstrate that MAC features well-positioned nucleosomes downstream of transcription start sites (TSS) likely connected with promoter proximal pausing of RNA polymerase II, as well as in exonic regions flanking both the 5′ and 3′ splice sites. In contrast, nucleosomes in MIC are more delocalized. Nucleosome occupancy in MAC and MIC are nonetheless highly correlated with each other and with predictions based upon DNA sequence features. Arrays of well-positioned nucleosomes are often correlated with GC content oscillations, suggesting significant contributions from cis-determinants. We propose that cis- and trans-determinants may coordinately accommodate some well-positioned nucleosomes with important functions, driven by a process in which positioned nucleosomes shape the mutational landscape of associated DNA sequences, while the DNA sequences in turn reinforce nucleosome positioning.