Hello Gordon,

I am using edgeR to do pairwise comparison for 3 conditions. The avg values for each condition are:

AVG_C1 : 1596

AVG_C2: 1354

AVG_R: 1934

C1 vs. R gives me logFC of -0.333

and C2 vs. R gives me logFC of 0.066

logFC of -0.33 indicates the gene is down regulated in C1 - which is correct looking at the avg values.

However, logFC of 0.066  indicates that the gene is up-regulated in C2 - which s NOT correct looking at the avg values.

I get correct sign for some genes, while others don't seem right. I am not worried about the magnitude.

I've checked my code, but not sure why there is discrepancy in the sign for logFC:

counts <- read.table( file = "juanCounts.txt" , header = TRUE, row.names=1)

A <- c("C2")

B <- c("R")

n <- as.numeric(4)

group <- c(rep(A, n) , rep(B, n))

counts<- data.frame(counts[,grep(paste(c(A,B),collapse="|"),names(counts))])

cds <- DGEList( counts, group = group )

cds <- calcNormFactors( cds )

cds <- estimateCommonDisp( cds )

cds <- estimateTagwiseDisp( cds )

de.cmn <- exactTest( cds ,  pair = c( A,B ) )

resultsTbl.cmn <- topTags( de.cmn , n = nrow( de.cmn$table ) )$table

Could you please advice if I missed something. If it's helpful I could send you the counts matrix.

Thanks for all the help.

Well, for starters: when you are estimating avg abundance counts by hand, I'm guessing you're using raw counts? Ie. counts that are not normalized by sequencing depth per sample?

The point I am trying to make is that the sign for the logFC doesn't seem right. 

Another example for C1 vs. R comparison is:

Gene1:

AVG_C1 : 18274

AVG_R: 2140

logFC is -0.35

Gene2:

AVG_C1 : 18791

AVG_R: 22027

logFC is -0.19

For Gene1 , avg C1 > avg R, and logFC sign is negative

For Gene2, avg C1 < avg R, and logFC sign is also negative.

If the sign is based on Condition 1 then shouldn't the logFC be positive for gene1 and negative for gene2.

Could someone help with this issue.