low R1/R2 (Illumina) overlapping

[LFMar]

Dear experts, I performed a paired-end amplicon (16S V3/V4 region) sequencing (2x250) on Miseq (Illumina) and the trimmed reads present a low (20%) overlapping between R1 and R2. I use this approach oftenly without problems. Any idea about what happened or how can I treat this data? Thanks in advance.

[nucacidhunter]

I guess quality of reads 3’ end (more likely Read2) was low so they have been trimmed and the length is not enough for overlap. You may try merging first and then filter low quality reads or lower trimming stringency.

[Brian Bushnell]

I guess quality of reads 3’ end (more likely Read2) was low so they have been trimmed and the length is not enough for overlap. You may try merging first and then filter low quality reads or lower trimming stringency.

Agreed. I think the best approach is usually to merge the raw (or adapter-trimmed) reads, then optionally try quality-trimming and merging the remaining unmerged pairs. BBMerge does this internally and it substantially improves the merge rate as opposed to only merging raw reads or only merging trimmed reads.

In cases like these it's also useful to explain for the forum what kind of data you have and how you did the trimming, by the way. Quite often people trim to extremely high levels like Q25, which is a bad idea, particularly for low-diversity amplicons which often have very low quality.

Depending on the dataset, it's also possible to error-correct rather than quality-trimming, or even extend non-overlapping reads so they overlap. This is mainly useful for randomly-sheared data, though, and not applicable to amplicons.