Hi,
I'm working on the DE analysis with DESeq2, here's my experimental design:
Genotype Time Replicates
Wild_type 6hr 3 for each
Wild_type 10hr
Wild_type 12hr
cht7 6hr
cht7 10hr
cht7 12hr
CHT7HA 6hr
CHT7HA 10hr
CHT7HA 12hr
With the following code, I'm able to generate the result tables comparing two different groups of genotype+time (e.g. Wild_type_6hr_vs_cht7_6hr):
Salmon_dds <-DESeqDataSetFromTximport(gene_quant,
colData = sampleTable,
design = ~genotype + time + time:genotype)
Salmon_dds$genotype <- factor(Salmon_dds$genotype, levels= c("Wild_type","cht7", "CHT7HA"))
Salmon_dds$time <- factor(Salmon_dds$time, levels = c("6","10","12"))
Salmon_dds$group <-factor(paste0(Salmon_dds$genotype, Salmon_dds$time))
design(Salmon_dds1) <- ~group
results(Salmon_dds, contrast = c("group", "cht76", "Wild_type6"), tidy = TRUE)
If I understand correctly, the resulting table can help me to identify the DE genes in cht7_6 compared to wild_type_6.
However, I also want to pick a specific group and compare it to all of the groups to find the DE genes, e.g. genes in cht7_6hr which have a greater than 2-fold expression difference (and padj <0.05) from their mean expression of all samples (BaseMean). Is there any way to generate a table like this?
The reason for doing this is to to get a list of DE genes which are differentially expressed in at least one group across all groups so I can do the follow-up gene clustering.
Thanks!
Well I have also recently did analysis on interaction type of design in DESeq2, what I would do in this case, I would use contrast and perform comparison keeping one condition as reference (as you are doing) and extract all the genes that are DE (log2FC >=1 || log2FC <=-1 && padj <=0.05) and combine the list. Further, I would extract normalized count value either in terms of rld or vst and perform gene clustering. This is just an suggestion, others may suggest better answer.
Thanks for the answer! It's a good idea to take one condition (e.g. wild type at 6hr) as the reference for others. The reason I wanted to do the one-vs-all (as described in my question) comparison is that there're a lot of papers using this method to get the DEGs for gene clustering, and they used DESeq2 to do that. Anyway, I'll try to get the DE genes as you suggested, thanks a lot!