16S Paired-End Illumina MiSeq Data in Qiime

Question

[Ruth M.]

I am learning to use Qiime to analyze 16S V3/V4 sequencing data from the Illumina Miseq (PE 2x300).

I'm wondering if there is any regular convention in deciding whether or not to pair ends before processing data. Because the second read from the Illumina 600v3 2x300 gives you poor quality after 200-250 bases, it seems that there would be many reads that don't get paired if using any kind of pairing program, right?

Is it possible to simply merge the two fastq files for read 1 and 2, or would that not work because of directionality?

I was going to use multiple_split_libraries_fastq.py, but if I have read 1 and 2 separate, that won't work either, correct?

Thanks in advance for any information!

Answer 1

[Brian Bushnell]

I suggest merging the reads with BBMerge (part of the BBMap package) prior to analysis. If the quality is so bad that you get a low merge rate, you can add the flag "loose" or "vloose" to increase it, at the expense of a slightly higher false-positive rate; or add "qtrim=r trimq=10" to do quality-trimming first. But overall, merged reads are easier to analyze and will have a lower error rate, so you should get more accurate clustering.

Answer 2

[Charles Warden]

1) If the read merging options in QIIME's join_paired_ends.py don't work well, I've found that PEAR usually works pretty good, typically merging a greater percentage of reads that still pass the QIIME quality score filter at a similar rate.

2) You can try using RTAX (inside or outside QIIME), which is supposed to work without merging the reads.

3) If you are still having problems, you can probably just use the forward read and get similar classification results.

Answer 3

[Picasa]

What I usually do is using trimmomatic to perform QC processing then merging the 2 PE with FLASh.

https://ccb.jhu.edu/software/FLASH/

http://www.usadellab.org/cms/?page=trimmomatic