Hi, I performed analysis using old 1.14 version of DESeq2. I had to repeat such analysis now with version 1.24. I've received very different results. With 1.14 I've received 63 hits of differentially expressed genes, while with 1.24 - 293 hits, and 18 overlapped. I try to understand where the difference comes from? Maybe I did sth wrong? for both I've used the followed command: conditions <- factor( c("P1", "P2", "P3", "P4", "P5", "P6", "CTRL", "CTRL", "CTRL", "CTRL", "CTRL")) conditions <- relevel(conditions, "CTRL") design <- data.frame( condition=conditions ) rownames(design) <- colnames(counts) dataset <- DESeqDataSetFromMatrix(countData = counts, colData = design, design = ~condition) dataset <- DESeq(dataset) deresults <- results(dataset) deresults <- deresults[ deresults$padj < 0.05 & complete.cases(deresults$padj), ] deresults <- deresults[ order(deresults$log2FoldChange, decreasing=T),] foldchange <- 2**deresults$log2FoldChange foldchangeinverse <- 1/foldchange newcolumns <- data.frame(GeneID=rownames(deresults), Foldchange=foldchange, Foldchangeinverse=foldchangeinverse) deresults <- cbind( newcolumns, deresults) write.table(as.data.frame(de_results), file="results.xls", sep="\t", quote=F, row.names=F)
Thanks for your answer. The code is below. Last night I've discovered quite strange behaviour. With absolutely the same counts matrix I'am receiving different number of hits of differentially expressesed genes depending on how I name condition. Namely: with this:
I've got more than 200 hits.
while with this:
I ve got about 50 hits.
Could you give me advise why is that? what should I change? I've used Deseq2 1.14, and with DEseq 1.26 was about 75 hits for first scenario and about 300 hits for the second. What is going on? I am not too much experienced (yet ;) in bioinformatics.
The full code is below.
See
levels(dds$condition)
And as I said previously check the specific contrast printed at the top of the results table when printed to the console.
The point of printing this contrast at the top of the table is so users can be reminded of what test they are performing.
Thanks, now I understand your point.
I have 5 controls and 6 patients to compare. Could you tell me what would be the correct approach:
or
The analysis is up to you. This support site is for software related questions. I’d recommend discussing your analysis plan with a statistician.
Thank you very much for your help!