Heavy-tailed prior distributions for sequence count data: removing the noise and preserving large differences

Abstract

Motivation In RNA-seq differential expression analysis, investigators aim to detect those genes with changes in expression level across conditions, despite technical and biological variability in the observations. A common task is to accurately estimate the effect size, often in terms of a logarithmic fold change (LFC). Results When the read counts are low or highly variable, the maximum likelihood estimates for the LFCs has high variance, leading to large estimates not representative of true differences, and poor ranking of genes by effect size. One approach is to introduce filtering thresholds and pseudocounts to exclude or moderate estimated LFCs. Filtering may result in a loss of genes from the analysis with true differences in expression, while pseudocounts provide a limited solution that must be adapted per dataset. Here, we propose the use of a heavy-tailed Cauchy prior distribution for effect sizes, which avoids the use of filter thresholds or pseudocounts. The proposed method, Approximate Posterior Estimation for generalized linear model, apeglm , has lower bias than previously proposed shrinkage estimators, while still reducing variance for those genes with little information for statistical inference. Availability and implementation The apeglm package is available as an R/Bioconductor package at https://bioconductor.org/packages/apeglm , and the methods can be called from within the DESeq2 software. Supplementary information Supplementary data are available at Bioinformatics online.

Histone H2AK119 Mono-Ubiquitination is Essential for Polycomb-Mediated Transcriptional Repression

“…Differential expression analyses were performed using the R package DESeq2 v1.20 (Love et al, 2014) using default parameters. Log2FoldChanges and adjusted p-values were corrected using the apeglm (Zhu et al, 2019) and IHW (Ignatiadis et al, 2016) packages, respectively. Genes with an absolute log2 fold change of 1.5 and FDR < 0.05 were considered as differentially expressed (Table S2).…”

Human pluripotent stem cell modeling of tuberous sclerosis complex reveals lineage-specific therapeutic vulnerabilities

“…DESeq2 analysis modelled fold change between time points using the cell line as an independent factor (using the model '~cell_line + condition'), and fold changes were calculated using lfcShrink function, applying the apeglm method (v 1.2.1) (Zhu et al, 2018). Gene ontology enrichment analysis was performed using clusterProfiler R package (Yu et al, 2012).…”

Polypyrimidine Tract Binding Proteins are essential for B cell development

“…5. RNA from 15,000 to 200,000 FrB cells was et al, 2014) by extracting differences between different genotypes in pair-wise comparisons using "apeglm" method as a shrinkage estimator (Zhu et al, 2019).…”

“…Read counts mapped over genes were counted with HTSeq (v0.10.0) (Anders et al, 2015) with f bam -r names no -a 10 parameters and the Mus_musculus.GRCm38.93.gtf annotation from Ensembl.Reads mapping to immunoglobulin genes (including V, D, J gene segments, light and heavy immunoglobulin genes and T cell receptor genes) were excluded before DESeq2 analysis. Differences in mRNA abundance were computed with DESeq2 (v1.22.1)(Love et al, 2014) by extracting differences between different genotypes in pair-wise comparisons using "apeglm" method as a shrinkage estimator(Zhu et al, 2019).…”