pca with VST of DESeq2 give 100% variance in PC1
1
0
Entering edit mode
@7e6a2774
Last seen 1 day ago
Italy

Hello, before doing the DESeq2 analysis on my bulk RNA-seq data, I used SVA to identify new covariates and I want to visualize my pca with vst command in DESeq2, so I did the following passage (I also delete the batch effect of my RIN for the visualization):


mm_ALpcrRD <- model.matrix(~ condition + RIN, colData(dds_ALpcrRD))
mm0_ALpcrRD <- model.matrix(~1, colData(dds_ALpcrRD))
fit_ALpcrRD <- svaseq(matrice_ALpCR_vs_ALRD, mod=mm_ALpcrRD, mod0=mm0_ALpcrRD)

dds_ALpcrRD$SV1 <- fit_ALpcrRD$sv[,1]
dds_ALpcrRD$SV2 <- fit_ALpcrRD$sv[,2]

design(dds_ALpcrRD) <- ~  condition + SV1+ SV2
vsd_ALpcrRD <- vst(dds_ALpcrRD, blind=FALSE)
covariates_ALpcrRD <- colData(vsd_ALpcrRD)[,c("SV1","SV2")]

assay(vsd_ALpcrRD) <- limma::removeBatchEffect(assay(vsd_ALpcrRD), batch = vsd_ALpcrRD$RIN,
                                             design=mm_ALpcrRD)

plotPCA(vsd_ALpcrRD,intgroup=c("condition"))

With this passage I have a pca wht 100% of the variance in PC1 and 0% in PC2, do you think it is possible? Is the first time that happen this thing to me. What does this can mean? Thank you!

DESeq2 sva • 144 views
ADD COMMENT
0
Entering edit mode
@james-w-macdonald-5106
Last seen 14 hours ago
United States

Perhaps Mike Love will be along later to comment on the order of vst and removeBatchEffect (not sure you should be using both here?), but anyway, you might consider either

A.) Not using RIN in your model, as the surrogate variables will probably do a better job of adjusting for RNA quality B.) If you do include RIN, do note that you are using it as an adjustment variable, in which case you should also include it in your null model.

0
Entering edit mode

Thank you for the answer. I used the same command on another matrix and I do not have this problem. I put the RIN in the model because, in another question, that I open here, you told me that it was the correct way to move, since, in this case I can find surrogate associate with it, in fact, in other case, my results are quite good.

ADD REPLY

Login before adding your answer.

Traffic: 445 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6