Super-delta2: an enhanced differential expression analysis procedure for multi-group comparisons of RNA-seq data

Abstract

Motivation We developed super-delta2, a differential gene expression analysis pipeline designed for multi-group comparisons for RNA-seq data. It includes a customized one-way ANOVA F-test and a post-hoc test for pairwise group comparisons; both are designed to work with a multivariate normalization procedure to reduce technical noise. It also includes a trimming procedure with bias-correction to obtain robust and approximately unbiased summary statistics used in these tests. We demonstrated the asymptotic applicability of super-delta2 to log-transformed read counts in RNA-seq data by large sample theory based on Negative Binomial Poisson (NBP) distribution. Results We compared super-delta2 with three commonly used RNA-seq data analysis methods: limma/voom, edgeR, and DESeq2 using both simulated and real datasets. In all three simulation settings, super-delta2 not only achieved the best overall statistical power, but also was the only method that controlled type I error at the nominal level. When applied to a breast cancer dataset to identify differential expression pattern associated with multiple pathologic stages, super-delta2 selected more enriched pathways than other methods, which are directly linked to the underlying biological condition (breast cancer). Conclusions By incorporating trimming and bias-correction in the normalization step, super-delta2 was able to achieve tight control of type I error. Because the hypothesis tests are based on asymptotic normal approximation of the NBP distribution, super-delta2 does not require computationally expensive iterative optimization procedures used by methods such as edgeR and DESeq2, which occasionally have convergence issues. Availability Our method is implemented in a R-package, “superdelta2”, freely available at: https://github.com/fhlsjs/superdelta2 Supplementary information Supplementary data are available at Bioinformatics online.

FastMix: a versatile data integration pipeline for cell type-specific biomarker inference

“…While raw RNA-seq reads are discrete, common pre-processing steps can make the RNA-seq data much more normal. Recent comparative studies (Cui et al, 2021;Law et al, 2014;Rapaport et al, 2013;Ritchie et al, 2015;Smyth et al, 2005) showed that DGEA designed for continuous data can achieve comparable, sometimes even slightly better, performance than those based on discrete models. We plan to extend FastMix based on generalized linear mixed-effects regression, so that it can integrate high-dimensional data that cannot be approximated by normal distributions.…”

“…Based on our previous work on gene expression normalization (Cui et al, 2021;Liu et al, 2017;Qiu et al, 2013Qiu et al, , 2014, we know that trimming can introduce noticeable bias in statistical inference. We derive the trimmed and bias-corrected covariance estimates for B as follows…”

“…After obtaining BT , we adopt a method in Appendix 1 of Cui et al (2021) to correct bias in b induced by trimming. Finally, random effects are re-computed with EBLUP using bias-corrected covariance and fixed effects.…”

IFIH1/IRF1/STAT1 promotes sepsis associated inflammatory lung injury via activating macrophage M1 polarization

“…In these multi-group comparisons, the Tukey’s style post-hoc pairwise test were utilized to analysis the high-throughput experiments and a false discovery rate ( FDR ) - adjusted P < 0.05 was set as the cut-off [18] , [21] . These high-throughput experiments included ChIP-seq data and differentially expressed analysis on mRNA microarray data of macrophage polarization (GSE46903).…”

Orthodenticle homeobox OTX1 is a potential prognostic biomarker for bladder cancer

“…RNA-seq data of 560 samples were covered in the transcriptome profiles for BLCA BC in TCGA database (TCGA-BLCA), containing 496 tumor tissues from patients with BC and 64 normal tissues from healthy volunteers. Edger & Limma/Voom were used to analyze differentially expressed genes in TCGA transcriptome data [ 19 ], and the conditions for screening differentially expressed genes were as following: | log (FC) | >1 & FDR < 0.05 or adj. p value <0.05.…”