edgeR logFC up/down regulation

[steve.eis]

Hey everyone,

i analysed my Illumina data with edgeR. I have to say that i am relativ new to bioinformatics and i find it hard, to decide, whether my genes are up or down regulated.
I comared two groups: 1h vs 0h
And now i want to know, if genes in 1h are up/down regulatet compared to 0h.
I used the following code to filter my results table:

et <- glmLRT(fit, coef=2)
dge=et$table
dge=subset(dge, PValue <= 0.05)
direction=vector()
for (i in 1:length(dge$logFC))
        {
                if (dge$logFC[i] > 0)
                        {direction[i] = c("UP")}
                else
                        {direction[i] = c("DOWN")}
        }
direction=as.factor(direction)

The Problem is, that i got the following results:
DOWN UP
372 1376

But there is also a bild in function to summarize the up/down regulation which brings:

summary(dt <- decideTestsDGE(et))

[,1]
-1 103
0 39645
1 198

Is it wrong that i said, that logFC >0 = UP? And i am not sure how the logFC rates are being calculated! Is it logFC= (1h/0h) when compared 1h vs 0h or the other way round???
I am really puzzeld and i hope someone can halp me!

[Thomas Doktor]

You should filter according to the FDR value and not the raw p-value, that's why you are seeing more differentially expressed genes using your own function compared to the built-in in edgeR. Why not simply use the edgeR function?

[steve.eis]

Thanks a lot for your fast reply!

Do you mean to use this function:

de <- decideTestsDGE(et, p=0.05, adjust="BH")

and if i do so by comparing 1h vs. 0h which ones are up/down regulated compared with what?

[Gordon Smyth]

Why not simply use topTags(et)?

To read about how logFC are calculated for glms, see Sections 3.2.3 and 3.2.4 of the edgeR User's Guide:

http://bioconductor.org/packages/release/bioc/vignettes/edgeR/inst/doc/edgeRUsersGuide.pdf