Differential splicing

limma can also detectgenes that how evidence of di_erential splicing between conditions. One can

test for di_erentialsplicing associated with any contrast for a linear model.

In this case, the matrixof counts should be at the exon level, with a row for each exon. For

example,

> dge <-DGEList(counts=counts)

> dge$genes$GeneID<- GeneID

where counts is amatrix of exon-level counts, and GeneID identi_es which gene each exon belongs

to. Then _lter andnormalize:

> A <-rowSums(dge$counts)

> dge <-dge[A>10,,keep.lib.sizes=FALSE]

> dge <-calcNormFactors(dge)

Then apply the voomtransformation and _t a linear model:

> v <-voom(dge,design,plot=TRUE)

> fit <-lmFit(v,design)

Now we can test fordi_erential splicing associated with any coe_cient in the linear model. First

run the di_Splicefunction:

> ex <- diffSplice(fit,geneid="GeneID")

Then

>topSplice(ex,coef=2,level="gene")

will _nd genes that showevidence of di_erential splicing associated with the second coe_cient in

the linear model. Theoutput is similar that from the limma topTable function. More detail can be

obtained by

72

>topSplice(ex,coef=2,level="exon")

which will showindividual exons that are enriched or depleted relative to other exons in thesame

gene. To display thepattern of exons in the top genes:

> plotSplice(ex)