teyden
commented
3 years ago My suspicion is that there is a bug in the example, and would love clarification from the authors. (Again thanks for doing this research)
From the example, it shows this:
otu.tab <- t(throat.otu.tab)
gmpr.size.factor <- GMPR(otu.tab)
dim(otu.tab)
[1] 856 60
otu.tab.norm <- t(t(otu.tab) / gmpr.size.factor)GMPR is computed on the OTU table where in otu.tab (836 OTUs x 60 samples), the rows are OTUs and columns are samples. But based on the documentation, the OTU matrix should have OTUs arranged in columns and samples in rows. When running the further steps, of producing a normalized OTU table following the example, many NA's are produced. And I believe this is because gmpr.size.factor has many NAs in it as the size factors are being computed on the OTUs when it should be computed on samples since library size is a feature of the samples.
Based on @BinhongLiu's comment, it seems you had caught this error and computed gmpr.size.factor <- GMPR(t(otu.tab)) instead which should be accurate, I think. Meaning the downstream computation should ideally work fine. I tried it and it worked for me, without producing any NA's for this example.
An OTU table matrix, where OTUs arranged in columns and samples in rows.
Hi, Thank you for developing this useful tool for microbiota data normalization! I'm trying to perform the demo data analysis, but there were some errors when I performed the analysis.
when perform the analysis through library:
when perform the analysis through source the function:
What should I deal with these NAs in the normalized table? Thank you! Hongbin liu