Entering edit mode
marika.catapano
•
0
@marikacatapano-15006
Last seen 6.2 years ago
Hi,
I have used two different versions of DESeq2 (1.16 and 1.18) and they give me completely different results loosing a lot of differentially expressed genes!
Since I want to compare different datasets and I will be analysing them over time how can I keep my analyses consistent? ( I can't just not update the packages, obv)
Thank you in advance.
Marika
HI Micheal,
Thanks for your prompt response and sorry for my delayed one.
I miss-typed the versions: it's 1.18 and 1.12.
Here my code:
WB_GPP_counts_cleaned.int = sapply(WB_GPP_counts_cleaned[1:16], as.integer)
countData<- as.data.frameWB_GPP_counts_cleaned.int)
columnNames<-c(colnames(countData))
condition <-c(rep("control", 7), rep("cases", 9))
sex <- c("Female", "Male", rep("Female",7), "Male", rep("Female",6))
genotype <- c(rep("_WT",7), rep("_IL36RN",4), rep("_WT",2), "_IL36RN", rep("_WT",2))
colData<-as.data.frame(cbind(columnNames, condition, sex, genotype))
colData$group <- factor(paste0(colData$condition, colData$genotype))
row.names(colData) <- colnames(countData)
dds1<-DESeqDataSetFromMatrix(countData, colData, design = ~ sex + group)
dds1$group <- relevel(dds1$group, ref = "control_WT")
dds1<-estimateSizeFactors(dds1)
dds<-DESeq(dds)
DEGs_genotype_GPPIL36RN_CTRLs <- as.data.frame(results(dds, contrast = c("group", "cases_IL36RN", "control_WT")))
Thanks in advance for your time on that
Version 1.12 is from April 2016, so two years out of date now. The big change was in moving LFC shrinkage from DESeq() to a subsequent function lfcShrink(). Every such change is made carefully after examining performance on benchmark datasets, and so when starting a new project you should just use the most up to date version of the package. If you’ve started with version 1.12 however and you need no changes in the list, you need to stick with that version. You can do so easily by obtaining the linked version of R and installing Bioconductor as you would normally.
That's what I suspected...
Thanks a lot!