mothur

Which commands require prior use of the "sub.sample" command?

#1

Hello,

I have been through the forum regarding the “sub.sample” command.

From what I understood, alpha and beta diversity indicators are calculated without prior use of the “sub.sample” command, but with specifying subsample=T (or subsample=xxx) in the commands “summary.single” and “summary.shared” / “dist.shared commands”, respectively. The latter command generates distance matrices that refer to a subsample of the whole dataset (right?), and such matrices are visualized using PCoA, NMDS or used in commands like AMOVA and HOMOVA. In addition, for the commands “corr.axes”, “get.communitytype”, “metastats” etc. one’s need to run the command “sub.sample” first.

Here are my questions:

  1. if i want to switch to PRIMER to generate NMDS graphs (instead of using MOTHUR , shame on me!), is it correct to run the “sub.sample” + “get.relabund” commands to generate a . relabund file that will imported in PRIMER?
  2. in the commands “heatmap.sim”, “heatmap.bin”, “venn”, “get.sharedotu” and “get.coremicrobiome”, should I use the “subsampled” dataset (generated with the “sub.sample” command), or the whole dataset ? I think it is the subsampled dataset, but I am not 100% sure.

Thanks for your help,
Isabelle

#2

Don’t subsample before running dist.shared, it will subsample before calculating the dissim matrix repeatedly then average so more robust.

i never use the commands in Q2, I do that sort of thing in R on the subsampled otu matrix.

Finally, please don’t use PCOA. It is only ever appropriate when the underlying gradient that your communities are reacting to is linear. I’ve never seen a microbial system that responds to change linearly. If you have found the rare system that does respond linearly, it’s probably appropriate to use PCA instead because if it’s linear it’s likely normal. Microbial communities are essentially never linear or normal, so use NMS

#3

Dear kmitchell,

Thank you for your fast reply.

I did not want to subsample before running dist.shared, but WHILE running dist.shared (dist.shared=xxx.shared, subsample=T). Does that sound correct ? I think this is the only way to get the subsampled OTU matrix used in the commands in Q2.

Regarding PCoA: indeed I will not use PCoA (I always use nMDS for microbial communities).

Kind regards,
Isabelle

PS: mistake in my previous post: “From what I understood, alpha and beta diversity indicators are calculated without prior use of the “sub.sample” command, but with specifying subsample=T (or SIZE=xxx) in the commands “summary.single” and “summary.shared” / “dist.shared commands”, respectively.”

1 Like
closed #4

This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.