Mismatch between PE reads

[tsangkl]

Hi all, I have 10 sets of illumina PE 250bp data.
The reads are 16S V3V4 regions.

The quality of reads drop dramatically especially in the 3'end.
I have trimmed the 3'end and merged the paired end reads using usearch7 fastq_mergepairs with 25 bp min overlap and max 1 mismatch.

Since the quality are bad and many reads are removed and some are too short to have overlapping region for merging.

But my major concern is the mismatch. I found that there are many mismatch between the forward and reverse reads in the overlapping region even I have trimmed the bad quality end. And fewer than half of the merged reads have exact match in the overlapping region.

Is it a problem?

[GenoMax]

You should consider first merging and then trimming as needed.

Someone else just reported that this is the correct way to do things: http://seqanswers.com/forums/showpos...52&postcount=9

[tsangkl]

Thank you GenoMax for your reply.

I have also done merging before trimming.
But there are more mismatch in the overlapping region if merge is done before trimming due to the bad quality at the 3'end.

If I limit the max no. of mismatch, a lot of (>70%) data were unnecessary removed.

So I am quite unconfident to use this data for further analysis

[Brian Bushnell]

If the quality is really terrible, I suggest merging, then trimming the unmerged reads, and then merging those. Don't worry about exact matches in the overlap region; part of the point of merging is to fix errors in the overlap region. It's the responsibility of the merger to ensure that the merges are correct, so as long as you are using a good program, it's inadvisable to cripple it by allowing at most 1 mismatch, particularly for such long reads.