DADA2 filterAndTrim with previously quality trimmed sequences. Is it advisable?

[auroralabastida]

Dear developer,

I am processing Illumina 2x250 paired-end reads corresponding to a 16S V3-V4 region sequencing. My usual pipeline implies primer and adapter elimination with Cutadapt followed by quality trimming with Trimmomatic. Here I eliminate any 5' and 3' bases with sequence quality lower than 3, apply a sliding-window filtering (window of size 10 with min. average quality of 15) and filter-out any surviving reads with average quality below 25 and length shorter than 50.

I am planning to use these pre-processed reads as input for DADA2 filterAndTrim, which I would use mainly to implement the Maximum Expected Errors filter and to eliminate any remaining Ns with parameters:

maxN=0, maxEE=c(2,2), truncLen = c(0,0), trimLeft = 0, trimRight=0, truncQ=0

I would then continue with the steps listed in the DADA2 Pipeline Tutorial (1.16)

Is it advisable to use quality-trimmed reads as input for dada2? Do you have any advice about the chosen filterAndTrim parameters?

Thanks in advance

[benjjneb]

In general you should not apply "sliding-window filtering" that cuts off reads at variable positions prior to using DADA2. In general, you should enforce a truncLen during the filterAndTrim step (outside of extremely length-variable loci like ITS).

Pre-processing to remove primers prior to DADA2 is fine, but in most use cases, you are best served in leaving the quality filtering to filterAndTrim using the standard maxEE and truncLen approaches.

[auroralabastida]

Thank you very much. I will certainly follow your recommendations. About the fw and rv read merging: Is it advisable to pre-merge the reads (ej. with FLASH) and then use them as single-end input for DADA2? I have seem several discussions elsewhere where this is either recommended or strongly discouraged.

[benjjneb]

Is it advisable to pre-merge the reads (ej. with FLASH) and then use them as single-end input for DADA2?

Not recommended.