Differential expressed genes in DESeq

Question

[Lila M]

Hi guys, I have an issue... I'm doing some analysis using DESEq2. By default I know that when I create the object dds using

dds <- DESeqDataSetFromMatrix(countData = count.table,
  colData = data.frame(treatment=treatment),
  design = ~ treatment)

an as treatment (control vs recurrence) by default R R will choose as reference level the factors based on alphabetical order, so in this case, the reference level would be control. But I have a question about this, for the downstream analysis, when I run for example:

dds <- DESeq(dds)
res <- results(dds)
res2Order <-res2[order(res2$pvalue), ]
write.table(res2Order, file="my _table")

I assume that the genes reported in my table with p-value < 0.05 are deferentially expressed in cotrol vs. recurred, but does it means that they are more or less expressed in control than recurred, but how can I know if this difference is based on a higher or lower expression? For example,can I confirm that a gene with p-value < 0.05 is more expressed in control than recurrence?

Many thanks in advance!

Answer 1

[Devon Ryan]

You'll note that there's a column with the fold change. Its sign dictates the direction of change.

[Lila M]

Thank you for the information, as is the first time that I have to deal with it, could you please help me to understand the results based on log2FoldChange?

ID                      log2FoldChange          pvalue
ENSG00000132002.6       -1.3096108594    6.38592315283661E-15
ENSG00000188517.13      2.568933441      1.05420723121764E-13
ENSG00000204388.6       1.4787338766     1.99686292186552E-12
ENSG00000240747.6       0.1054737261     1.53129979101136E-11

So in this specific case, does it means that the gene ENSG00000132002.6 is differential expressed in control vs recurred samples and indeed is less expressed in controls? Is the same for the log2fold change values ~ 0?

Many thanks!

[Nicolas Rosewick]

Yes , as Devon explain the sign of the log2FC tells you the direction of change. In the case of ENSG00000132002.6 the log2FC is -1.3096108594, thus foldchange = 2^-1.3096108594 = 0.4034297 so 2.478747 (1/0.4034297) times less expressed.

[Lila M]

But is less expressed in recurred? I understand that if I've included as treatment: control and recurred, and by default R will choose as reference level the factors based on alphabetical order, control should be my reference level and in that case, recurred is less expressed than control. Is this assumption ok?

[Nicolas Rosewick]

Yes. You can check the normalized read counts for ENSG00000132002 to be sure.

norm_count <- counts(dds,normalized=T)

[Lila M]

Ok, many thanks for this tip!!!