Question

rlog, lfcShrink and vst on gene subset

1

Entering edit mode

giulia.pasquesi ▴ 10

@giuliapasquesi-15143

Last seen 6.5 years ago

I am analyzing a dataset that includes both gene and transposable element raw counts.
I am only interested in analyzing the differential expression of TEs, but still need to create and normalize the entire dataset. For the DE analysis itself it is not a major problem, because I can retrieve the results and filter out the genes. However, I was interested in plotting the variance-stabilized transformation of the data (or the rlog) and the PCA for the TEs only, ignoring the genes.
Is there a way to do this directly in DeSeq2?
I was thinking of providing an additional feature data column specifying "gene" or "TE", but then I don't know if it is possible post-normalization to use only the "TE" subset to make transformations and lfcShrink.

Thank you so much

Giulia

deseq2 subsetting plot • 2.0k views

ADD COMMENT • link updated 6.5 years ago by Michael Love 43k • written 6.5 years ago by giulia.pasquesi ▴ 10

0

Entering edit mode

Have you tried simply row-subsetting the DESeqDataSet before calling the model fitting functions?

ADD REPLY • link 6.5 years ago Wolfgang Huber ★ 13k

score 0 · Answer 1 · 2018-06-11

0

Entering edit mode

Michael Love 43k

@mikelove

Last seen 3 days ago

United States

If you want to plot for example, the PCA for just some samples, you can provide a subsetted DESeqTransform object to plotPCA:

plotPCA(vsd[idx,])

ADD COMMENT • link 6.5 years ago Michael Love 43k

0

Entering edit mode

I did't know about the [idx] flag but isn't it gonna specify the samples I want to include? I would like to plot the variances for only a subset of genes (therefore by rows and not by column in a matrix. Provided the normalization was based on all genes, otherwise the library size would be incorrect and bring to a weird normalization) for all samples.

I apologize if I wasn't clear the first time.

ADD REPLY • link 6.5 years ago giulia.pasquesi ▴ 10

1

Entering edit mode

The way indexing works in R is, if you use [ ... , ... ], the first element indexes rows (here, genes) and the second element indexes columns (samples). If you leave out an element, it provides all. So [idx,] gives a subset of the rows and all of the columns.

See here:

http://bioconductor.org/packages/release/bioc/vignettes/SummarizedExperiment/inst/doc/SummarizedExperiment.html#anatomy-of-a-summarizedexperiment