DESeq2 very high p-values and padj

Question

[madkitty]

I'm comparing a control vs Treatment A and control vs.Treatment B with DESeq2 though the pvalues and padj are really high. Most padj are 0.99 or so, and most pvalues range anywhere between 0.1 to 0.9. I'm suspecting taht the way I wrote the chunk of code below doesn't compare well C vs A and C vs B, what's wrong in the following code?..

# DESeq1 libraries
library( "DESeq2" )
library("Biobase")

r12 = read.csv("20140807_R12.csv", header = TRUE, row.names=1)

samples <- data.frame(row.names=c("C", "A", "B"), condition=as.factor(c("C", "A", "B")))
samples$condition <- relevel(samples$condition, "C")

dds <- DESeqDataSetFromMatrix(countData = as.matrix(r12), colData=samples, design=~condition)
dds$condition <- factor(dds$condition, levels =c("C", "A", "B"))

dds <- DESeq(dds)
res <- results(dds)
head(res)

DataFrame with 6 rows and 6 columns
             baseMean log2FoldChange     lfcSE       stat    pvalue      padj
            <numeric>      <numeric> <numeric>  <numeric> <numeric> <numeric>
3BHSD        1.946289    -0.13137862 0.3349360 -0.3922499 0.6948736 0.9963381
5_8S_rRNA   25.628958     0.05835918 0.4797943  0.1216338 0.9031891 0.9963381
5HT2A        2.748903    -0.08777809 0.3950686 -0.2221844 0.8241703 0.9963381
5S_rRNA   1179.943635     0.39496852 0.4627465  0.8535312 0.3933648 0.9963381
7SK       1325.918947     0.39695551 0.4538413  0.8746571 0.3817605 0.9963381
A1BG         5.639654    -0.28433104 0.4541317 -0.6260982 0.5312506 0.9963381

[Mikael Huss]

You are just looking at the first few genes of res. What do you get from e g

res.sig <- res[res$padj < 0.1,]

or

head(res[order(res$padj),])

?

[madkitty]

I've looked at the whole spreadsheet and there's no padj under 0.1......

[Devon Ryan]

Do you really have no replicates (that's what it looks like at least)?

[madkitty]

Indeed, we have no replicate for these samples. I don't have that issue when I run DESeq2 with replicates and multiple treatments. Is there a fix for that? ...

[Nicolas Rosewick]

If you don't have any replicate it's normal that p-values are high

[madkitty]

Looks like it ..

Answer 1

[Devon Ryan]

There's no fix because there's nothing to fix. I'm a little surprised you got adjusted p-values that low even. There are no meaningful p-values for your experiment because computing them would be impossible. No replicates means no meaningful p-values.