Paired End Data Merging: Before Or After ?

Question

[Aytaç]

Hey Everyone. I want to perform a metagenomic project. I have paired end data in fastq format. I checked forums and many other blogs including this one. Merging paired end data should be done before quality controls (Trimming, Adaptor removing, Chimera removing) or after?

I even asked this question to chatGPT but it gave me skeptical answers. It told me to do merging after the quality filtering but before chimera removing.

In a forum (link : https://drive5.com/usearch/manual/pipe_readprep_merge.html) it says do merging before quality filtering.

In addition i will use Kraken2 for this job and i know Kraken2 can deal with paired end reads but it just stucked in my head. Which order is best for these steps:

x - Quality improving (Removing low quality(phread) score reads) x - Chimera Removing x - Paired end data merging x - Adaptor removing

Answer 1

[Mark]

I would not merge. I think historically it was done to improve read quality and improve denovo assembly or alignment (i.e. that muscle link). Nowadays with the improvement in technology, its not needed. The exception is maybe if you anticipate major read (R1/R2) overlapping significantly because the target fragments are small.

I would also suggest very VERY soft trimming. Soft trimming is defined as removing low quality reads (< 15 phred score), sliding window trimming where the average score is <15. Do not do hard trimming (remove X bases from left and right).

If you are using kraken, I highly suggest you read this protocol from the authors: https://www.nature.com/articles/s41596-022-00738-y for best practices. They have extensively published about the misuse of kraken. Follow their advice!

[Aytaç]

Thank you very much sir. I will look at the article you sent.

I got adaptor sequences from the data's article so I will be only removing them to move on.