Low overlap due to poor reverse read quality in 2×300 bp V3–V4 amplicon sequencing (DADA2)

General Discussion
1

Hello,

I am planning to conduct 16S rRNA amplicon sequencing targeting the V3-V4 region (~460 bp) using Illumina Miseq i100 (2x300 bp paired-end) for rhizosphere samples.

In theory, this setup should provide sufficient overlap of ~140 bp (= 600 bp - 140 bp), which is well above the Illumina's recommended minimum of ~ 50bp.

However, in practice, I'm observing that the 3' end of the reverse reads has significantly lower quality. After quality trimming, the effective overlap region often becomes shorter than expected, leading to poor merging efficiency or loss of reads.

I'm currently using DADA2 with the following filtering parameters:

  • filterAndTrim: truncLen = c(250, 220), maxEE = c(2, 5), truncQ = 2

With these settings, the total retained length is ~470 bp, which leaves only ~10 bp of overlap for a ~460 bp amplicon—below the recommended minimum for reliable merging. This seems to explain the merging issues I’m seeing.

I'm wondering:

  • Is this a common issue with 2x300 bp Illumina runs for V3-V4 amplicons?

  • How do you typically handle low-quality reverse reads in this context?

  • Do you relax trimming parameters, trim more aggressively, or adjust merging criteria?

  • Is it bettern to prioritize overlap (keeping reads longer) or quality (more aggressive filtering)?

  • For rhizosphere microbiome studies, would you recommend switching to a shorter region (e.g., V5-V7) to improve merging efficiency?

Any insights or practival recommendations would be appreciated.

Thanks!

2

User Support ello and welcome to the forum!

Thank you for providing many details.

Yes, I see it pretty often with the V3-V4 region.

First, I try all the options available to recover paired reads. If nothing works, I discard reverse reads and go for "forward only" as a last resort.

I try to choose the minimum trimming parameters that still ensure sufficient overlap, decrease the minimum overlapping region (to 6), and allow for more errors in the reads.

I prioritize overlap to avoid bias towards bacteria with shorter V3-V4 regions, assuming that sequencing errors are more randomly distributed.

I don't have a suitable benchmark, but in a study performed on the pig fecal microbiome V1-V2 and V4 regions performed as well as V3-V4. I also used V1-V2 for rhizosphere microbiome, but I didn't compare it to V3-V4. Honestly, I prefer V1-V2 (2x250) over V3-V4, exactly because of situations like yours.

Best,

1 Like