Hello
I'm trying to reproduce the analysis done in this paper and I've tried R's siggenes and Tmev, but I don't get the same list of differentially-expressed genes using both methods, even when tuning the delta parameter, nor do I get the two genes (RPV1 and IFI30) he picked for qPCR from any of my lists of differentially-expressed genes. I'm pretty sure I chose the wrong samples. I tried two-class unpaired SAM, for control vs. chron's disease, but even that doesn't include the genes he chose. I also tried multiclass SAM, still no good.
The author used such samples:
GSM's: normal, crohn's disease, ulcerative colitis, infectious colitis, indeterminate colitis
Then he grouped them as:
control, unaffected and affected (Note: the affected/ unaffected come from the fact that data were taken from symptomatic vs asymptomatic patients having the diseases)
Do I use multiclass or two-class SAM: "control vs affected" or "control vs unaffected" or "affected vs unaffected".
In the paper he said he used healthy vs diseased..
Grateful if someone could enlighten me.
I'm going to try the options meanwhile.
Did you try contacting the author?
I'll try a bit more before. I thought that would be my last resort before contacting the author. But thanks anyways. Will do so if I don't manage.
Link to the paper: http://www.ncbi.nlm.nih.gov/pubmed/21849664
I already have the paper, but there's no mention of how the samples were compared by SAM analysis.. I thought it would be easy, like comparing 2 samples.. I've done it like thus "control vs affected", "control vs unaffected" and "affected vs unaffected") with R, but I only managed to get only one of the mentioned genes (control vs affected, with a delta of 2.3; FDR 0.008; 10.7 false calls vs 489 called genes)..There's probably something I've missed..
Is multiclass comparison (haven't tried it yet) meaningful in this case. (I'm new to this kind of analysis, but it's so interesting) Thank you.