Purchase this article with an account.
Sudeep Mehrotra, Revital Bronstein, Daniel Navarro-Gomez, Ayellet Segre, Eric A Pierce; Evaluating methods for differential gene expression and alternative splicing using internal synthetic controls. Invest. Ophthalmol. Vis. Sci. 2019;60(9):3437.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Emerging evidence suggests that many genetically unsolved inherited retinal disease (IRD) cases may be due to mutations affecting differential gene expression (DGE), differential transcript/isoform expression (DTE), or altered splicing (AS) of known IRD disease genes. Detecting these regulatory changes demands a diverse set of bioinformatic analyses. In this work we present results from assessing and comparing a number of widely used methods for detecting DGE, DTE and AS using internal control ‘spike-in’ sequences (Sequins) in RNA-seq data.
Sequins are full-length spliced mRNA isoforms that bear no resemblance to the mouse or human reference genomes. The Sequins were added in known concentrations in separate mixes in each replicate for case and control samples. Our choice of software programs for analysis was based on the following factors: widely used and vetted in the research community, continuous support from authors and compatibility with high performace compute clusters. For evaluation, we performed sensitivity and specificity analysis and computed the correlation of determination (R2). Read alignment was performed using STAR, featureCount was used for gene expression and RSEM and Kallisto for isoform expression quantification. For DGE, DESeq2 and edgeR, and Sleuth, EBSeq and edgeR for DTE were used. Finally, we tested JunctionSeq and MAJIQ for AS detection.
The STAR aligner detected all synthetic genes, with intron level sensitivity ≤ 84%. RSEM and Kallisto showed detection of isoform expression, with R2 ≤ 0.92 and 0.95, respectively. Both DESeq2 and edgeR showed high correlation R2 ≤ 0.90. This was dependent on depth of coverage and number of replicates. Sleuth, EBSeq and edgeR showed poor sensitivity and specificity for DTE. Both JunctionSeq and MAJIQ showed sub-optimal sensitivity (60%-70%).
Tools for RNA-seq read alignment and detection of DGE performed reasonably. More efforts are needed to improve specificity and sensitivity of DTE and AS detection. Low expression of genes/isoforms accompanied with sequencing depth and limited number of replicates does impact sensitivity and specificity of each tool. Inclusion of internal controls in RNA-seq experiments allows accurate determination of detection levels, and better assessment of DGE, DTE and AS accuracy.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.
This PDF is available to Subscribers Only