Hello,

I have an experiments where i have 6 groups. I have followed the documentation in order to compare pairwise groups.

I have normalized the data as follows

... p = cumNormStat(css) css = cumNorm(css, p = p) pd <- pData(css) mod = model.matrix(~treatment,data=pd) colnames(mod) = levels(pd$treatment) # Apply fitZig function settings = zigControl(maxit = 10, verbose = TRUE) res = fitZig(obj = css, mod = mod, control = settings) zigFit = res$fit finalMod = res$fit$design > head(finalMod) GROUP_A0 GROUP_A14 GROUP_A28 GROUP_B0 GROUP_B14 GROUP_B28 scalingFactor Sample1 1 0 0 0 0 0 0.4038131 Sample2 1 0 0 1 0 0 0.4222330 Sample3 1 0 0 0 0 0 0.4499575 Sample4 1 0 0 1 0 0 0.4179200 Sample5 1 0 0 1 0 0 0.7259592 Sample6 1 0 0 0 0 0 0.6535187

# Then i create a pairwise foldchanges comparison as follows

contrast.matrix = makeContrasts(GROUP_A14 - GROUP_A0, levels = finalMod) > contrast.matrix Contrasts Levels GROUP_A14 - GROUP_A0 GROUP_A0 -1 GROUP_A14 1 GROUP_A28 0 GROUP_B0 0 GROUP_B14 0 GROUP_B28 0 scalingFactor 0 fit2 = contrasts.fit(zigFit, contrast.matrix) fit2 = eBayes(fit2) top <- topTable(fit2,number=Inf)

Let´s look at a particular OTU (OTIU_320) for from the top table:

OTUname logFC AveExpr t P.Value adj.P.Val B Kingdom 70 OTU_320 -7.448524 9.953013 -8.102799 2.586277e-12 3.004845e-11 15.91786 Bacteria Phylum Class Order Family Genus 70 Firmicutes Bacilli Lactobacillales Streptococcaceae Streptococcus

**It seems that OTU_320 is downregulated in GROUP_A14 compared to GROUP_A0, at least -7.44 times (log2). **

Now I have extracted all values using the MRcounts (css, log=F) and selected athe same specific OTU and computed the mean for each group

OTU_320:

the mean value for GROUP_A0 is 2900.35263 (log2 = 11.5) for this particular OTU

the mean value for GROUP_A14 is 5067.54602 (log2=12.3) for this particular OTU

Just by looking at the logration there is a difference of 12.3-11.8 = 0.8

**So (please correct me if i´m wrong i expect an upregulation of this particular OTU in GROUP_A14 of 0.8 log2).**

**i don´t understand why there is a so big difference (downregulation of 7.44 when using the recommended procedure vs up regulation of 0.8 when comparing means of normalized counts) and would like to understand if this is the result of the fitZig algorithm ? Can you help me on this ? I need to understand this to be able to explain to the biologists that play on their side with the normalized counts.**

**By the way if i compare average expression there are also different (9.95 in toptable) vs 12 by comparing normalized data means for this two groups. **

Thanks,