Many ASVs missing in different runs

[kreetelyll]

Hi!
I did the denoising separately as supposed to but now I ran into a new issue when analyzing the data. I have data from 3 timepoints and what I could see is that many ASVs (on the picture still named OTU, sorry for that ) from the timepoint 1 do not exist in other timepoints (so that ASV from timepoint 2 or 3 is not existent in timepoint 1 and vice versa). On the other hand, ASVs from 2nd and 3rd timepoint overlap rather nicely. When looking in R and adding taxonomy I could see that the different ASVs still count as the same taxonomix unit (all the way down to species) in the clusters.
My question is why are the ASVs different although they are the same bacteria and does it come from denoising step with DADA2 where all 3 runs were run separately? Furthermore, is it somehow possible to have ASVs be named the same if they represent the same bacteria. Our statistician likes to do some analyzes on OTU/ASV level so that is why it would be nice to work it out.
I hope my question was understandable.

module load python/3.6-QIIME2-2019.7

 qiime dada2 denoise-single --p-n-threads 40 --i-demultiplexed-seqs run1-sequences.qza --p-trunc-len 130 --p-trim-left 0 --o-representative-sequences rep-seqs-run1.qza --o-table table-run1.qza --o-denoising-stats stats-run1.qza

qiime dada2 denoise-single --p-n-threads 40 --i-demultiplexed-seqs run2-sequences.qza --p-trunc-len 121 --p-trim-left 0 --o-representative-sequences rep-seqs-run2.qza --o-table table-run2.qza --o-denoising-stats stats-run2.qza
 
qiime dada2 denoise-single --p-n-threads 40 --i-demultiplexed-seqs run3-sequences.qza --p-trunc-len 121 --p-trim-left 0 --o-representative-sequences rep-seqs-run3.qza --o-table table-run3.qza --o-denoising-stats stats-run3.qza

qiime feature-table merge --i-tables table-run1.qza --i-tables table-run2.qza --i-tables table-run3.qza --o-merged-table table-merged.qza

qiime feature-table merge-seqs --i-data rep-seqs-run1.qza --i-data rep-seqs-run2.qza --i-data rep-seqs-run3.qza --o-merged-data rep-seqs-merged.qza

Taxonomy (I have V4 region):

qiime feature-classifier classify-sklearn --i-classifier silva-132-99-515-806-nb-classifier.qza --i-reads rep-seqs-merged.qza --o-classification taxonomy.qza

[thermokarst (Matthew Ryan Dillon)]

The reason these are so different between runs is because you have not used the same trim and trunc settings for all runs. Because of that, you have completely different ASVs in each run. You will need to decide on settings that apply well to all three runs in order to be able to merge them. Another alternative (that I don’t recommend unless it is not possible for you to re-run), is to perform closed-reference OTU clustering prior to merging.