Observed ASV richness drops after adding 3rd MiSeq library — why do I lose ~100 species when I merge libraries?

[Basubi]

I have 3 Illumina MiSeq libraries (same region, same prep).
When I process libraries individually I get ~350 fungal ASVs and ~300 bacterial ASVs (observed ASVs).
I merged library A + B → richness stays ~350 (fungi) / ~300 (bacteria).
But when I merge library C with A+B the observed ASV richness drops for both kingdoms (I lose ≈100 ASVs for bacteria and ≈100 ASVs for fungi — final counts ~200 bacteria, ~250 fungi).
Surprisingly, library C when it is analyzed separately the number of ASVs increased ~100 more that when it is merged with the other two libraries.
What could cause richness to decrease when adding a third library?

[timanix (Timur Yergaliyev)]

Welcome to the forum!

@Basubi

I merged library A + B → richness stays ~350 (fungi) / ~300 (bacteria).
But when I merge library C with A+B the observed ASV richness drops for both kingdoms (I lose ≈100 ASVs for bacteria and ≈100 ASVs for fungi — final counts ~200 bacteria, ~250 fungi).

Are you merging the outputs of dada2, or do you first merge the libraries and then run dada2?

Multiple sequencing runs merged together before DADA2 may mess up the error training step and lead to biases in some libraries. It is why it is recommended:

• run all libraries separately, but with identical parameters
• merge feature tables and representative sequences after dada2

Let us know if you already implemented it like I have described above and still facing the same issue - then we’ll need to figure out what is going on there.

Best,

[Basubi]

Yes. I merged those tables after dada2.

[timanix (Timur Yergaliyev)]

That’s strange.

Did you perform filtering based on the relative abundance or prevalence before calculating diversity metrics? We definitely need more data to solve this issue!