Entering edit mode
6.5 years ago
BrINClHOF
•
0
HI all,
Does anybody have any suggestions on R packages for rarefaction creation? I have about 34,000 different species from 17 different samples I want to test for richness, any other suggestions would be greatly appreciated. Thank you in advance!
unrelated of the incompleteness of your question, better not use rarefaction
Would that be because of the overall size of the dataset? If so would breaking it down into smaller subsets representing each of the 17 soil samples be any better?
Rarefaction is an applicable normalization method in metagenomics, but Paul McMurdie and Susan Holmes show in their work how it can skew your results and hence suggest to use better suiting normalization methods adapted from the EdgeR and DeSEQ2 packages. mastal511 mentioned vegan and qiime which I believe adopted those methods in the meantime, definitely qiime2. No need to split to subsets, fmpov.
Have you tried the vegan package in R? You can also do rarefaction with qiime.
I have not tried vegan yet, but I used qiime2 to generate all of my data would I have to run everything back through the pipeline to do the rarefaction with qiime?
samples of what? A sample, to me is a tube of serum, how do you get more species than samples?
what do you need from "test for richness" ? Sounds quite rich.
They are 17 soil samples which were tested for their microbial diversity and across all 17 gave 34,000 different OTU's. As far as testing for richness I'm just wanting to visualize how diverse the soil samples were
Okay maybe a Venn Diagram can show how many things are shared or unique between sets. But you have 17 groups, so that's a terrible mess. If it was 2-3 groups this could be a good visualization. With 17 groups, do you have any kind of experimental hypothesis? "Showing how diverse" could just be a 17-wide bar graph, if you can deal with overlaps and uniqueness somehow.