Skip to contents

Perform differential expression analysis

Source: R/differential-expression.R
deAnalysis.Rd

performMirnaDE() and performGeneDE() are two functions provided by MIRit to conduct miRNA and gene differential expression analysis, respectively. In particular, these functions allow the user to compute differential expression through different methods, namely edgeR (Quasi-Likelihood framework), DESeq2, limma-voom and limma. Data deriving from NGS experiments and microarray technology are all suitable for these functions. For precise indications about how to use these functions, please refer to the details section.

Usage

performMirnaDE(
  mirnaObj,
  group,
  contrast,
  design,
  method = "edgeR",
  logFC = 0,
  pCutoff = 0.05,
  pAdjustment = "fdr",
  filterByExpr.args = list(),
  calcNormFactors.args = list(),
  estimateDisp.args = list(robust = TRUE),
  glmQLFit.args = list(),
  glmQLFTest.args = list(),
  DESeq.args = list(),
  useVoomWithQualityWeights = TRUE,
  voom.args = list(),
  lmFit.args = list(),
  eBayes.args = list(),
  useArrayWeights = TRUE,
  useWsva = FALSE,
  wsva.args = list(),
  arrayWeights.args = list(),
  useDuplicateCorrelation = FALSE,
  correlationBlockVariable = NULL,
  duplicateCorrelation.args = list()
)

performGeneDE(
  mirnaObj,
  group,
  contrast,
  design,
  method = "edgeR",
  logFC = 0,
  pCutoff = 0.05,
  pAdjustment = "fdr",
  filterByExpr.args = list(),
  calcNormFactors.args = list(),
  estimateDisp.args = list(robust = TRUE),
  glmQLFit.args = list(),
  glmQLFTest.args = list(),
  DESeq.args = list(),
  useVoomWithQualityWeights = TRUE,
  voom.args = list(),
  lmFit.args = list(),
  eBayes.args = list(),
  useArrayWeights = TRUE,
  useWsva = FALSE,
  wsva.args = list(),
  arrayWeights.args = list(),
  useDuplicateCorrelation = FALSE,
  correlationBlockVariable = NULL,
  duplicateCorrelation.args = list()
)

Arguments

mirnaObj

A MirnaExperiment object containing miRNA and gene data

group

The variable of interest for differential expression analysis. It must be the column name of a variable present in the metadata (colData) of a MirnaExperiment object. See the details section for additional information

contrast

A character object that specifies the groups to be compared during differential expression analysis, separated by a dash (e.g. 'disease-healthy'). Note that reference group must be the last one, for additional information see the details section

design

An R formula that indicates the model to fit. It must include the variable of interest (group) together with eventual covariates (e.g. '~ 0 + disease + sex'). Please note that group variable must be the first one. See the details section for additional information

method

The statistical package used to compute differential expression. For NGS experiments, it must be one of edgeR (default), DESeq2, and voom (for limma-voom). Instead, for microarray data, only limma can be used

logFC

The minimum log2 fold change required to consider a gene as differentially expressed. Optional, default is 0

pCutoff

The adjusted p-value cutoff to use for statistical significance. The default value is 0.05

pAdjustment

The p-value correction method for multiple testing. It must be one of: fdr (default), BH, none, holm, hochberg, hommel, bonferroni, BY

filterByExpr.args

A list object containing additional arguments passed to edgeR::filterByExpr() function. It is used when method is set to edgeR or voom

calcNormFactors.args

A list object containing additional arguments passed to edgeR::calcNormFactors() function. It is used when method is set to edgeR or voom

estimateDisp.args

A list object containing additional arguments passed to edgeR::estimateDisp() function. It is used when method is set to edgeR. Default is list(robust = TRUE) to use the robust parameter

glmQLFit.args

A list object containing additional arguments passed to edgeR::glmQLFit() function. It is used when method is set to edgeR

glmQLFTest.args

A list object containing additional arguments passed to edgeR::glmQLFTest() function. It is used when method is set to edgeR

DESeq.args

A list object containing additional arguments passed to DESeq2::DESeq() function. It is used when method is set to DESeq

useVoomWithQualityWeights

Logical, whether to use the limma::voomWithQualityWeights() function or just the limma::voom() function. It is used when method is set to voom. Default is TRUE

voom.args

A list object containing additional arguments passed to limma::voom() function or limma::voomWithQualityWeights() function. It is used when method is set to voom

lmFit.args

A list object containing additional arguments passed to limma::lmFit() function. It is used when method is set to voom or limma

eBayes.args

A list object containing additional arguments passed to limma::eBayes() function. It is used when method is set to voom or limma

useArrayWeights

Logical, whether to use the limma::arrayWeights() function or not. It is used when method is set to limma. Default is TRUE

useWsva

Logical, whether to use the limma::wsva() function or not. It is used when method is set to limma. Default is FALSE

wsva.args

A list object containing additional arguments passed to limma::wsva() function. It is used when method is set to limma

arrayWeights.args

A list object containing additional arguments passed to limma::arrayWeights() function. It is used when method is set to limma

useDuplicateCorrelation

Logical, whether to use the limma::duplicateCorrelation() function or not. It is used when method is set to limma. Default is FALSE

correlationBlockVariable

It is the blocking variable to use for limma::duplicateCorrelation(). Default is NULL

duplicateCorrelation.args

A list object containing additional arguments passed to limma::duplicateCorrelation() function. It is used when method is set to limma

Value

A MirnaExperiment object containing differential expression results. To access these results, the user may run the mirnaDE() and geneDE() functions for miRNAs and genes, respectively.

Details

When performing differential expression for NGS experiments, count matrices are detected and method parameter must be one of edgeR, DESeq2, and voom. On the other hand, when dealing with microarray studies, only limma can be used.

To calculate differential expression, MIRit must be informed about the variable of interest and the desired contrast. In particular, the group parameter must be the name of a variable present in the metadata (colData) of a MirnaExperiment object, which specifies the variable used to compute differential expression analysis, between the groups indicated in contrast. Specifically, contrast must be a character vector that defines the levels to compare separated by a dash. For example, if we have a variable named 'condition', with two levels, namely 'disease' and 'healthy', we can identify differentially expressed genes in 'disease' samples compared to 'healthy' subjects by specifying: group = 'condition' and contrast = 'disease-healthy'. Furthermore, the user needs to specify the model to fit expression values. To do so, the user has to state the model formula in the design parameter. Please note that for a correct inner working of these functions, the group variable of interest must be the first variable in model formula. Moreover, the user can include in the design any other sources of variation by specifying covariates that will be taken into account. For instance, if we want to compare 'disease' subjects against 'healthy' individuals, without the influence of sex differences, we may specify design = ~ condition + sex, where 'sex' is also a variable present in the metadata (colData) of mirnaObj.

Notably, for all the methods available, the user can supply additional arguments to the functions implemented in edgeR, DESeq2 and limma. Therefore, the user has finer control over how the differential expression analysis is performed. In this regard, for microarray studies, the user may opt to include weighted surrogate variable analysis (WSVA) to correct for unknown sources of variation (useWsva = TRUE). Moreover, for microarray data, the arrayWeights() function in limma can be used to assess differential expression with respect to array qualities. Also, the duplicateCorrelation() function in limma may be included in the pipeline in order to block the effect of correlated samples. To do this, the user must set useDuplicateCorrelation = TRUE, and must specify the blocking variable through the correlationBlockVariable parameter. Additionally, when using limma-voom, the user may estimate voom transformation with or without quality weights (by specifying useVoomWithQualityWeights = TRUE).

Functions

  • performMirnaDE(): Perform differential expression analysis for miRNAs

  • performGeneDE(): Perform differential expression analysis for genes

References

Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, Smyth GK (2015). “limma powers differential expression analyses for RNA-sequencing and microarray studies.” Nucleic Acids Research, 43(7), e47. doi:10.1093/nar/gkv007.

Law, CW, Chen, Y, Shi, W, and Smyth, GK (2014). "Voom: precision weights unlock linear model analysis tools for RNA-seq read counts". Genome Biology 15, R29

Robinson MD, McCarthy DJ, Smyth GK (2010). “edgeR: a Bioconductor package for differential expression analysis of digital gene expression data.” Bioinformatics, 26(1), 139-140. doi:10.1093/bioinformatics/btp616.

Love MI, Huber W, Anders S (2014). “Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2.” Genome Biology, 15, 550. doi:10.1186/s13059-014-0550-8.

Author

Jacopo Ronchi, jacopo.ronchi@unimib.it

Examples

# \donttest{
# load example MirnaExperiment object
obj <- loadExamples()

# perform miRNA DE with edgeR
obj <- performMirnaDE(obj,
    group = "disease", contrast = "PTC-NTH",
    design = ~ 0 + disease + patient, method = "edgeR"
)
#> Performing differential expression analysis with edgeR...
#> Differential expression analysis reported 61 significant miRNAs with p < 0.05 (correction: fdr). You can use the 'mirnaDE()' function to access results.

# perform miRNA DE with DESeq2
obj <- performMirnaDE(obj,
    group = "disease", contrast = "PTC-NTH",
    design = ~ 0 + disease + patient, method = "DESeq2"
)
#> Performing differential expression analysis with DESeq2...
#> Warning: some variables in design formula are characters, converting to factors
#> estimating size factors
#> estimating dispersions
#> gene-wise dispersion estimates
#> mean-dispersion relationship
#> final dispersion estimates
#> fitting model and testing
#> Differential expression analysis reported 87 significant miRNAs with p < 0.05 (correction: fdr). You can use the 'mirnaDE()' function to access results.

# perform miRNA DE with limma-voom
obj <- performMirnaDE(obj,
    group = "disease", contrast = "PTC-NTH",
    design = ~ 0 + disease + patient, method = "voom"
)
#> Performing differential expression analysis with voom...
#> Differential expression analysis reported 55 significant miRNAs with p < 0.05 (correction: fdr). You can use the 'mirnaDE()' function to access results.
# }