DESeq2 for multiple groups

Question

[banaalkhayrat123]

Hi everyone,

I am performing differential expression analysis using featureCounts files that I previously analyzed in UseGalaxy. However, I am now very confused. Does DESeq2 in UseGalaxy allow for comparisons such as (Overexpression cell line1 - Control1) - (Overexpression cell line2 - Control2)? I have two different cell lines, each with its own control, and I want to compare them. From what I understand, DESeq2 uses the first factor as the reference. What if I have two reference groups? In this case, would Limma-voom be a better option?

Another question: If I have two or three treatments but only one reference (control) and want to compare the treatments with each other, how should I set up the analysis? All treatments come from the same cell line, but they are part of different experiments, all sharing the same control. For example, I have cell line 1 with treatment 1, treatment 2, treatment 3, and a common control. How can I properly compare the treatments with each other?

Thanks in advance, Bana

Answer 1

[ATpoint]

For the first question: In DESeq2 I think you need to use a list with the contrast argument in results() to test fold changes of fold changes. I usually do this in limma so others might correct me if this is not accurate.

Here is a toy example:

library(DESeq2)

# Intercept-less (~0) design to make all groups accessible as coefficients in resultsNames()
set.seed(1)
dds <- makeExampleDESeqDataSet(n = 5000, m = 20, )
dds$group <- factor(rep(LETTERS[1:4], each = 5))
design(dds) <- ~ 0 + group
dds <- DESeq(dds)
res <- results(dds, contrast = list(c("groupA", "groupB"), c("groupC", "groupD")))

# Plot the one with smallest pvalue
gene_to_plot <- rownames(res[which(res$pvalue == min(res$pvalue, na.rm = TRUE)),])

data <- plotCounts(dds = dds, gene = gene_to_plot, intgroup = "group", returnData = TRUE)

library(tidyverse)
library(ggpubr)
data %>% mutate(membership = if_else(group %in% c("A", "B"), "ONE", "TWO")) %>%
  ggplot(aes(x = group, y = count, color = membership)) +
  geom_point(position = position_jitter(.2, 0, 0), size = 3) +
  stat_summary(fun = mean, geom = "crossbar", size = 1)

Here is a plot assuming A and B are treatment and control of cell line 1 and C and D of cell line 2.

For the second question, if you have the same control then you can directly compare the groups, like A-B, B-C, A-C. Does this make sense?

[banaalkhayrat123]

Thank you so much @ATpoint. That is amazing since I will try to run the analysis in RStudio. However, my question was namely in usegalaxy.eu platform, whether I can use the deseq2 tool in this platform to perform DEGs from multiple groups while taking into account the control group.

[banaalkhayrat123]

Thank you so much @ATpoint. That is amazing since I will try to run the analysis in RStudio. However, my question was namely in usegalaxy.eu platform, whether I can use the deseq2 tool in this platform to perform DEGs from multiple groups while taking into account the control group.

[ATpoint]

This you have to ask at the Galaxy support, they have a forum like this one here afaik.

[GenoMax]

https://help.galaxyproject.org/c/usegalaxy-eu-support/6 is the forum for Galaxy EU Support.

[banaalkhayrat123]

Yes Thank you!