DESeq2 coefficient names
1
0
Entering edit mode
jkhudyakov • 0
@jkhudyakov-23010
Last seen 2 days ago
United States

I am having trouble getting specific coefficient names after running DESeq2. The coefficients I get when I run resultsNames(dds) are "Treat1", "Treat2", etc instead of "Treat_B_vs_A", "Treat_C_vs_A", etc. Has anyone else had this issue? The code that I was running and the output are below:

samples$Treat <- factor(samples$Treat, levels = c("A","B","C","D","E","F"))
samples$Treat # A A A B B B C C C D D D E E E F F F # Levels: A B C D E F dds <- DESeqDataSetFromTximport(txi, colData = samples, design = ~ Subject + Treat) keep <- rowSums(counts(dds) >= 10) >= 3 dds <- dds[keep,] dds$Treat <- relevel(dds$Treat, ref = "A") dds <- DESeq(dds) resultsNames(dds) # "Intercept" "Subject1" "Subject2" "Treat1" "Treat2" "Treat3" "Treat4" "Treat5" sessionInfo( ) # R version 3.6.0 (2019-04-26) # Platform: x86_64-apple-darwin15.6.0 (64-bit) # Running under: macOS 10.15.7  Thank you! resultsNames DESeq2 coefficient • 152 views ADD COMMENT 0 Entering edit mode @mikelove Last seen 15 hours ago United States Looks strange. And Subject is a simple factor also? What about class(samples)? ADD COMMENT 0 Entering edit mode Hi Michael, Yes, Subject is also a simple factor. samples$Subject returns:

12 14 15 12 14 15 12 14 15 12 14 15 12 14 15 12 14 15 12 14 15 12 14 15 Levels: 12 14 15

class(samples) returns

"data.frame"

The strange thing is that I had it working properly last week, but it stopped giving specific coefficients when I ran it this week. I am not sure what I did differently - nothing I don't think.

0
Entering edit mode

Check if your installation is still ok: BiocManager::valid()

0
Entering edit mode

I get:

Bioconductor version '3.9'

• 115 packages out-of-date
• 0 packages too new

Versions of Bioconductor packages used for the analysis: tximport_1.12.3 tximportData_1.12.0 DESeq2_1.24.0 apeglm_1.6.0 IHW_1.12.0 rhdf5_2.28.1

I did the package update recommended by Bioconductor and re-ran DESeq2 but am still getting the weird coefficient names.

0
Entering edit mode

I'm not sure, could you try with the same data on a different machine?

Also what do you get with:

model.matrix(~Treat, data=samples)

0
Entering edit mode

I have 8 conditions with 3 biol reps each. I releveled the treatment factor so that every sample is compared to A (control).

As a (slight) aside, how would I specify the comparison B vs C if Treat2 is B vs A and Treat 2 is C vs A? Is this correct? contrast = c("Treat", "Treat3", "Treat2")

A_rep1 1 -1 -1 -1 -1 -1 -1 -1

A_rep2 1 -1 -1 -1 -1 -1 -1 -1

A_rep3 1 -1 -1 -1 -1 -1 -1 -1

B_rep1 1 1 -1 -1 -1 -1 -1 -1

B_rep2 1 1 -1 -1 -1 -1 -1 -1

B_rep3 1 1 -1 -1 -1 -1 -1 -1

C_rep1 1 0 2 -1 -1 -1 -1 -1

C_rep2 1 0 2 -1 -1 -1 -1 -1

C_rep3 1 0 2 -1 -1 -1 -1 -1

D_rep1 1 0 0 3 -1 -1 -1 -1

D_rep2 1 0 0 3 -1 -1 -1 -1

D_rep3 1 0 0 3 -1 -1 -1 -1

E_rep1 1 0 0 0 4 -1 -1 -1

E_rep2 1 0 0 0 4 -1 -1 -1

E_rep3 1 0 0 0 4 -1 -1 -1

F_rep1 1 0 0 0 0 5 -1 -1

F_rep2 1 0 0 0 0 5 -1 -1

F_rep3 1 0 0 0 0 5 -1 -1

G_rep1 1 0 0 0 0 0 6 -1

G_rep2 1 0 0 0 0 0 6 -1

G_rep3 1 0 0 0 0 0 6 -1

H_rep1 1 0 0 0 0 0 0 7

H_rep2 1 0 0 0 0 0 0 7

H_rep3 1 0 0 0 0 0 0 7

attr(,"assign")

[1] 0 1 1 1 1 1 1 1

attr(,"contrasts")

attr(,"contrasts")\$Treat

[1] "contr.helmert"

0
Entering edit mode

I see -- the issue is that you've set a non-default setting for contrasts in your .Rprofile.

I'd recommend to take that out if you want default contrasts from DESeq2.

0
Entering edit mode

Ahh that's right, I used helmert contrasts for a different project and didn't realize it changed it in the .Rprofile. I set it back to default and now it works. Mystery solved! Thanks so much for your prompt help and for writing these tools!