Dear developers,
Just short bug report.
While viewing table.qzv (in this example it is the Fecal Transplantation exercise) under the "Interactive Sample Detail" tab, I was playing with the rarefaction depth. I noticed that the "retained sequences" percentage is lowest at both ends. Please find attached two snapshots with the scroll at too ends of the scale.
Let me know if you need further details here,
Good luck,
Arnon
Thanks for reporting @arnon!
Perhaps I misunderstand you, but this looks consistent to me. The key point is that any samples with < X sequences are dropped, where X = the rarefaction depth.
You rarefy at 61 sequence per sample, and you get a total of 61 * 121 samples = 7381 sequences.
You rarefy at 8,374 sequences per sample and only 1 sample exceeds that depth, so you get a total of 8374 sequences.
So this makes sense: few sequences are retained at low and high rarefaction depths because 1. low rarefaction samples very few sequences and 2. high rarefaction samples a higher number of sequences but you are losing 99% of your samples in the process!
Sorry Nick, you are totally right. I was looking at the "Sampling Depth" more as the minimum cut-off value of filtering by library size (e.g. we take all the samples with higher depth) so 61 for example should retain most of the sequences. My confusion...
All the best.
A
I have a short follow up question on this. It now (qiime2-amplicon-2024.5, Linux) reports on "X retained features" instead of "sequences" in the older version and the first post in this topic.
isn't that a bit misleading as the term "features" is used for the actual amplicon variants/units, because in the first tab the actual number of features is reported
If in the first tab there are 3.8k features, doesn't make sense to retain 65k features with a sampling depth of 65k. Or am I mistaken, what was the reason behind the change? Thank you.
Hi @lxsteiner
I definitely see your point and have opened an issue regarding this on our github. Table Summary should say "Number of Unique Features" · Issue #313 · qiime2/q2-feature-table · GitHub
Just to clarify, feature is a purposely broad word. It could mean ASV, OTU, sequences, reads, metabolite data, etc..
So it is being used correctly in both places once to mean ASVs and once to mean reads but I do see how this is confusing esp. if you are thinking you only have 3.8K feaures but then you retain 20* more than that.
If you would like to follow along on progress with this issue. Check out the GH issue linked above!
