I'm running SAM from the R samr package to in multiclass mode to determine genes differentially expressed in one or more classes, and I'm struggling to interpret the output.
I'm running it on a 5000 row by 47 sample expression matrix (either RNA-seq count or batch-corrected log-transformed expression data) and all 5000 are showing as upregulated whichever method I'm using (SAM or SAMseq).
However, the columns for the classes "color.ind.for.multi.1" "color.ind.for.multi.2" "color.ind.for.multi.3" "color.ind.for.multi.4" in some cases are all zero, which according to the manual mean that the difference in expression in these four classes is not significantly higher than the grand mean (at 95% CI).
So my question is: why are all genes showing as upregulated? And why are some genes showing as zero significant changes in classes but still showing as upregulated?
Alternatively, is there a better way to do multi-class differential expression from RNA-seq data?
