Rarefy alpha beta metrics p sampling depth

Hello everyone,

I'm here to ask a question about rarefaction when performing core metrics. I used the lowest read value of my 97 samples which is 25,000 as --p-sampling depth parameter as I saw in these pages Microbiome/16s_qiime2_pipeline.md at master · WeersmaLabIBD/Microbiome · GitHub and workshop--06--quality_control.utf8 .
My questions are:

  • I did not rarefy my asvs table, to avoid losing important information as described in several publications. Knowing that I rarefied at a depth that corresponds to the smallest of my samples when I saw on the feature table.qzv, I have this message ''Retained 2,565,359 (38.15%) features in 97 samples at the specified sampling depth" which I don't really understand...
  • Another question, why do we have to rarefy at a minimal sequencing depth? is it just to be able to perform the core metric? since I kept all the 97 samples? Which means that I don't lose data?

Thank you very much

Rarefaction for now is a requirement when calculating core-metrics in Qiime2.
By choosing 25000 as a sampling depth, all samples were rarefied to this number.
But I would not worry - 25000 is a lot of sequences. You can create a alpha - rarefaction plot to check if it is sufficient (I think it is ok as it is).
If you choose a sampling depth higher than number of sequences in some samples, plugin will discard all these samples. But again, I am sure that 25000 reads is enough.

1 Like

Thank you for your answer @timanix, I have juste an other question this sentence: Retained 2,565,359 (38.15%) features in 97 samples at the specified sampling depth. I don't understand why it retained 38.15% feature since I have rarefied for core-metrics at the minimum sampling depth ?

Because in all other samples you have more reads that 25000. When rarefied, from each sample 25000 reads were randomly subsampled.
You didn't lose any samples, but overall number of sequence frequencies decreased.


Thank you very much for the clarification!

1 Like

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.