Suggestion regarding DESeq design of experiment

Question

[pixie@bioinfo]

Hello, I need some help regarding this group of RNA-seq data sets on different stress conditions (say A, B and C). The authors who actually performed these experiments performed cluster analysis using FPKM and normalized across all the stress conditions. In the paper, they mentioned they have kept all the experimental conditions same and varied only the stress. I am not interested in DEG analysis as of now, I only want to export the vst-transformed data for downstream analysis (cluster, co-expression network analysis, etc). How do I modify the code ? Kindly help .

Data:

Stress A

Ctrl_A_rep1
Ctrl_A_rep2
Str_A_1hr_rep1
Str_A_1hr_rep2
Str_A_6hr_rep1
Str_A_6hr_rep2

Stress B

Ctrl_B_rep1
Ctrl_B_rep2
Str_B_1hr_rep1
Str_B_1hr_rep2
Str_B_6hr_rep1
Str_B_6hr_rep2

Stress C

Ctrl_Crep1
Ctrl_C_rep2
Str_C_1hr_rep1
Str_C_1hr_rep2
Str_C_6hr_rep1
Str_C_6hr_rep2

R code I tried as of now:

library(DESeq2)
count_table<-read.table("all_expr.txt", header=T, row.names=1, sep="\t")
exp_design<-data.frame(row.names=colnames(count_table), condition = c("ctrl_A",  "ctrl_A", "A_1hr", "A_1hr", "A_6hr", "A_6hr",   "ctrl_B",  "ctrl_B", "B_1hr", "B_1hr", "B_6hr", "B_6hr",  "ctrl_C",  "ctrl_C", "C_1hr", "C_1hr", "C_6hr", "C_6hr"))

conditions=exp_design$condition
ds <- DESeqDataSetFromMatrix(countData=count_table, exp_design, formula(~0+condition))

Answer 1

[grant.hovhannisyan]

You should run the DESeq analysis and then,

data_norm<-rlog(dds)
rlog_data<-assay(all_data_norm)

Here I do rlog transformation, but you can do vst if you need the software to be faster.

[pixie@bioinfo]

So the design is fine or should there be any other factors considered ?

[grant.hovhannisyan]

Ok, I thought you are interested only in how to retrieve the normalized and transformed counts. But if you are not interested in DE analysis, then DESeq2 count normalization does not depend on design formulae. See here https://support.bioconductor.org/p/106548/ So if you only need normalized and transformed counts, then you can proceed further.

[pixie@bioinfo]

Okay thats great ..thanks