Skip to main content

Table 1 Analysis methods with their associated R packages and details concerning their implementation

From: A comparison of methods for multiple degree of freedom testing in repeated measures RNA-sequencing experiments

Method

R-Package(s)

Multiple DF Test

Library Adjustment Offset

Details

DESeq2

DESeq2 [6]

LRT

DESeq2 size factors

Default settings used, correlation ignored.

DESeq2 ∗

DESeq2 [6]

LRT

DESeq2 size factors

Default settings used, subjects treated as fixed effects to account for correlation.

edgeR

edgeR [4]

LRT

TMM offset

Default settings used, correlation ignored.

edgeR ∗

edgeR [4]

LRT

TMM offset

Default settings used, subjects treated as fixed effects to account for correlation.

limma

limma [9, 23]

Moderated F-test

NA

Count data transformed using the voom function. The duplicateCorrelation function was used with subject id as a blocking factor to account for correlation.

GEE

Custom R Functions, geepack [27]

Wald χ2 Test

DESeq2 size factors

Models fit using exchangeable working correlation structure. For small sample estimators, custom functions were created by modifying code from the geesmv package [28] to make it compatible with geepack and enable the use of offsets.

LMM

lmerTest [29]

F-test

NA

Data transformed using the variance stabilizing transformation from the DESeq2 package.

NBMM-AGQ

GLMMadaptive [30]

LRT

DESeq2 size factors

Model fit using the mixed_model function with a negative binomial distribution. Default settings used for all other parameters.

NBMM-LP

glmmADMB [31]

LRT

DESeq2 size factors

Model fit using the glmmadmb function with a negative binomial distribution. Default settings used for all other parameters.

NBMM-PL

Custom R Function

LRT

DESeq2 size factors

Custom function was created drawing from the glmm.nb function in the NBZIMM package [21]. Function was created to be compatible with the lmerTest package [29] in order for Satterthwaite degrees of freedom to be calculated.

rmRNAseq

rmRNAseq [18]

Moderated F-statistic with bootstrapped p-values

NA

Model fit using the TC_CAR1 function with the default parameters.