Help me find the bug
1
0
Entering edit mode
Michael • 0
@e828643d
Last seen 6 months ago
United States

I am trying to compare treatment (A) vs my control (C). A heavily downregulates genes, but for some reason my volcano plot shows that it is upregulated when compared to the control. The samples were not swapped (I've checked) and I got the correct distribution (downregulation) once, but since then have gotten the reverse. What is going wrong?

dds <- DESeqDataSetFromMatrix(countData = my_cts,
                          colData = my_coldata,
                          design = ~ condition)
dds

#prefilter row sums with less than 10 reads
keep <- rowSums(counts(dds)) >= 10
dds <- dds[keep,]

#factor levels
my_coldata$condition <- factor(my_coldata$condition, levels = c("C","A"))

#do i need these?
    #dds <- estimateDispersions(dds)
    #dds <- nbinomWaldTest(dds) 

#run DESeq2
dds <- DESeq(dds)

#check coefficients estimated by DEseq
resultsNames(dds)

#obtain results
res <- results(dds, contrast=c("condition","A","C")) #order is treatment, numerator, denominator

#shrink LFC
resLFC <- lfcShrink(dds, coef = 2, type = "apeglm") #coef is set to "condition A vs C"

#make volcano plot (the result that looks flipped) 
par(mfrow=c(1,1))
# Make a basic volcano plot
with(res, plot(log2FoldChange, -log10(pvalue), pch=20, main="Volcano plot", xlim=c(-4,4)))
with(subset(res, padj<.01 ), points(log2FoldChange, -log10(pvalue), pch=20, col="blue"))

output

Any advice is much appreciated.

DESeq2 HELP • 1.3k views
ADD COMMENT
1
Entering edit mode
@mikelove
Last seen 1 hour ago
United States

A very similar question was asked recently on the support site.

When you pass an object to a constructor function, it makes a copy. So later when you modify it, it doesn't affect the dds.

Here

#factor levels
my_coldata$condition <- factor(my_coldata$condition, levels = c("C","A"))

You've modified my_coldata but since dds has a copy, you haven't modified dds.

ADD COMMENT
0
Entering edit mode

Hi Michael, so does this mean I have to factor dds$condition as well? I though I could get the same result by running: res <- results(dds, contrast=c("condition","A","C"))

If not, how do I modify the code so that dds is modified properly?

ADD REPLY
0
Entering edit mode

You can fix my_coldata above the construction of dds.

Or you can change it once it is a column of the colData:

dds$condition <- doSomethingTo(dds$condition)

Specifying the two levels in contrast also works. You should verify that you are sure of the levels and their associated counts by looking at plotCounts for individual genes. Sometimes users find that samples are mislabeled by examining plotCounts.

ADD REPLY
0
Entering edit mode

Thanks! I checked my dds$condition before running factor and it looked like it had levels. Just to clarify, is this the correct order of events?

  1. factor(coldata)*
    1. make dds matrix
    2. prefilter and factor(dds)
    3. run DESeq
    4. results(dds, contrast = "")

is step 1 even necessary if we're factoring downstream? Also if I were to relevel dds to set a different reference after step 5, would I need to run deseq again and factor after and should this be saved to a different variable like dds2 because the model changes?

Plotting the individual genes helped by the way--it confirmed what I was seeing and it ended up being that my PI had given me the data in the wrong order lol...

ADD REPLY
1
Entering edit mode

Again, you can factor before making dds, or you can directly factor the column as part of dds. Either works. You don’t need to do both.

Yes if you change levels you’d generally need to re-run the GLM for apeglm to work, this is discussed in the vignette under shrinkage estimators.

ADD REPLY
0
Entering edit mode

Thank you so much!

ADD REPLY

Login before adding your answer.

Traffic: 725 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