how to get variation of each component in function plotPCA(DEseq)?

[xuenjun1]

Hello all,

I have the question regarding to the attached figure in this question: http://seqanswers.com/forums/showthr...hlight=plotPCA

How to get those percentage from each component? I have used the DESeq function plotPCA.

Here is the simplified code that I used:
cdsFullBlind = estimateDispersions( cdsFull, method = "blind" )
vsdFull = varianceStabilizingTransformation( cdsFullBlind )
print(plotPCA(vsdFull, intgroup=c("condition", "genotype")))

Thanks a lot!

[dpryan]

From the code for plotPCA:

rv = rowVars(exprs(x))
  select = order(rv, decreasing=TRUE)[seq_len(ntop)]
  pca = prcomp(t(exprs(x)[select,]))

ntop defaults to 500 and "x" would be vsdFull in your example.

[xuenjun1]

From the code for plotPCA:

rv = rowVars(exprs(x))
  select = order(rv, decreasing=TRUE)[seq_len(ntop)]
  pca = prcomp(t(exprs(x)[select,]))

ntop defaults to 500 and "x" would be vsdFull in your example.

Thanks a lot!

I just run and got this result
summary(pca)
Importance of components:
PC1 PC2 PC3 PC4 PC5 PC6 PC7 PC8
Standard deviation 91.3001 52.8806 26.28177 14.80544 12.35189 10.68246 9.35059 8.70695
Proportion of Variance 0.6469 0.2170 0.05361 0.01701 0.01184 0.00886 0.00679 0.00588
Cumulative Proportion 0.6469 0.8639 0.91755 0.93456 0.94640 0.95526 0.96204 0.96793
PC9 PC10 PC11 PC12 PC13 PC14 PC15 PC16 PC17
Standard deviation 7.61892 7.45351 7.31581 7.00614 6.63956 6.42668 6.28243 6.10567 5.91024
Proportion of Variance 0.00451 0.00431 0.00415 0.00381 0.00342 0.00321 0.00306 0.00289 0.00271
Cumulative Proportion 0.97243 0.97674 0.98090 0.98471 0.98813 0.99133 0.99440 0.99729 1.00000
PC18
Standard deviation 5.395e-14
Proportion of Variance 0.000e+00
Cumulative Proportion 1.000e+00

I guess the Proportion of Variance of PC1 and PC2 are the percentage that I want? And one more question, is the plotPCA output PC1 and PC2 in the attached figure corresponding to the first two components out of 18 PCs? Thanks a lot again.

[dpryan]

Yes, the proportion of the variance is what you want. PC1 and PC2 are from pca$x (the summary function you used doesn't show you everything in the pca object).

[dvanic]

Just as a note for other who may come across this post, I tried to use the above code with DESeq2 version 1.6.1, and got an error of

rv = rowVars(exprs(vsd))
Error in (function (classes, fdef, mtable) :
unable to find an inherited method for function ‘exprs’ for signature ‘"SummarizedExperiment

And at least for the version of DESeq2 I am using, the expression is actually obtained by using:

rv = rowVars(assay(vsd)

Then everything else works as expected.

/Posting this in case someone else google-finds this post and then gets the error I did/