increased sampling depth decreases observed features for some samples in rarefaction plot


I've created a rarefaction plot using following command:

qiime diversity alpha-rarefaction \
  --i-table table-dada2-GH_exp23_FW_200_fun.qza \ #filtered table for fungi
  --i-phylogeny rooted-tree_GH_exp23_FW_200_fun.qza \ #based on filtered seqs for fungi
  --p-max-depth 145509 \ #max number of reads for a sample
  --m-metadata-file metadata_file_GenomeQuebec2.csv \
  --o-visualization rare_faction_GH_exp23_FW_200_fun.qzv

and obtain following rarefaction plot:

as you can see, for some samples, the observed_features decreases when the sequencing depth increases. This puzzles me, as normally this would always increase?

Not sure if this matters, but I filtered my original dataset using:

#===========filter table for fungi only==========#

qiime taxa filter-table \
  --i-table table-dada2-GH_exp23_FW_200.qza \
  --i-taxonomy taxonomy_GH_exp23_FW_200.qza \
  --p-include Fungi  \
  --o-filtered-table table-dada2-GH_exp23_FW_200_fun.qza
#filter sequences for fungi only
qiime taxa filter-seqs \
  --i-sequences rep-seqs-dada2-GH_exp23_FW_200.qza \
  --i-taxonomy taxonomy_GH_exp23_FW_200.qza \
  --p-include Fungi  \
  --o-filtered-sequences seqs-dada2-GH_exp23_FW_200_fun.qza

What could be the reason for this decrease?

using qiime2-amplicon-2024.2

1 Like

Hi @Rob_DNA,
This definitely seems weird ...

Seems like you could have gotten some outlier random sampling? Could you try to re-run alpha-rarefaction and let me know if that fixes this?

Hi @cherman2 ,

the result is exactly the same.. but I fortunately found the issue .... :slight_smile:

I have DNA and cDNA of the same sample. So in the metadata column " sample" , the cDNA and DNA variants of the same sample, have the same identifier.

giving them a unique name solved the issue... silly mistake :wink: