We have a set of data, that by analyzing them in different scenarios the reported ASVs number is different. For example, if we run all the samples which are from different years together, we get totally different results comparing to the scenario if we run every year sample separately.
I am using DADA2 for Denosing in all the cases and the same classifier (V6-V8 bacteria).
Could it be because of the Maximum samples depth ?
Have you ever had such a case?
Do you know what impacts the number of ASV picking?
When you run DADA2, you need to work with samples on the same sequencing run, not with any other metadata category. The algorithm assumes it’s applied per-sequencing run, that no quality filtering has been applied, and that reads are single-end. If those assumptions don’t work for you or your experiment, Deblur is a good alternative.
My recommendation is to denoise your samples once using a fixed set of parameters. They don’t have to be perfect, but they need to be consistent, especially in terms of final length.
AAAG and AAAGG will be two different ASVs, even though if you’d trimmed them both to 4 nt they’d be the same.
Then, if you’re applying a rarefied alpha diversity metric, you need to use a consistent rarefaction depth. Rarefaction depth will also impact the number of observed features because they’re capped by that depth. (Run alpha-rarefaction on your data and test it for yourself!)
In general, I find it a lot easier to process all my data at once (feature table, taxonomic classification, rarefaction, alpha, and beta diversity) and then filter it later to get exactly what I want. But, if I process everything together, I know my parameters are consistent.