Is dereplication necessary?

[tsangkl]

Hi all,

I am new to 16S metagenomics,
now I have a batch of 16S illumina PE 250bp reads generated from 10 samples.
The reads target on V3 V4 region of 16S and around 440bp after merging the forward and reverse reads.

I have quality trimmed and merged the paired end reads,
I am now not sure whether I should dereplicate the reads before removing chimeric reads and map to greengenes 97% otu.
some said remove identical reads is needed due to artificial replicates.

Any advise on dereplicate issue?
I wonder if dereplication would alter the abundance counting on each otu?

Thanks!

[Brian Bushnell]

Dereplication would certainly alter the abundance count... if you are interested in quantifying coverage for any reason, do not dereplicate!

For this kind of dataset, there's no reason to dereplicate other than as a step before tree-building. But rather than dereplication I would suggest clustering prior to tree-building, and using the consensus (or a single representative) of each cluster to build the tree.

[tsangkl]

Thank you Brian.