Limma Voom Differential Splicing diffSplice() error

[Anomilie]

Hi,

I've been following the user's guide for Limma voom to perform differential splicing analysis and I have encountered a few problems.

Firstly, In chapter 15.4 the following line of code produces and error:

dge <- dge[A>10,,keep.lib.sizes=FALSE]
Error in object[[a]][i, , drop = FALSE] : incorrect number of dimensions

If I perform filtering using the following code which I believe does something similar

 
isexpr <- rowSums(dge$counts) > 10
x <- x[isexpr,]
dge <- calcNormFactors(x)

And then continue with the following as stated in the documentation I get an error:

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

ex <- diffSplice(fit, geneid="GeneID")
Error: could not find function "diffSplice"

Is there something (perhaps very stupid) that I'm doing wrong??

Thanks in advance

[Cytosine]

Please provide sessionInfo() so we can better assist you...

dge <- dge[A>10[COLOR="Red"],,[/COLOR]keep.lib.sizes=FALSE]
Error in object[[a]][i, , drop = FALSE] : incorrect number of dimensions

I see a double "," in there. That's causing your number of dimension error.

Error: could not find function "diffSplice"

Update Limma if you have an old version.

[Anomilie]

The double quote is directly copied from the documentation and is apparently requirement for subsetting the DGEList object.

the following code sorted out that issue (in case someone else has similar problems)

dge <- DGEList(counts=raw_data, genes=geneID)

In regards to the diffSplice error, the error message seems fairly obvious now, I had to update limma to version 3.20.

Thanks for your reply Cytosine

While on the topic of diffSplice, would you know the statistical difference between using that function and performing the DE with the standard limma voom pipeline using eBayes(fit)?

[Gordon Smyth]

While on the topic of diffSplice, would you know the statistical difference between using that function and performing the DE with the standard limma voom pipeline using eBayes(fit)?

It is exactly equivalent to a voom eBayes pipeline, but adapted to assess differential splicing instead of differential expression.