Number of reads after joining

[Tien Thi Thuy NGUYEN]

Dear everyone,

I have a question about fastq-join. Please help me to make it clear. Any help is appreciated so much. 

I have 10 reference fastq files downloaded from SRA. These files were originated from samples which were sequenced with V3-V4 region of 16S rRNA gene by Miseq Illumina platform.

 I tried to join these files by using command "fastq-join" in QIIME 1.9.1. 

As a result, the number of reads was around 5000 reads in each sample with parameter = 10. 

I feel confused. I think sequences outputted from Miseq Illumina sequencer have more than 10.000 reads. 

I don't know why, in this case, the number of reads was too low. 

In comparison with other fastq files that also sequences with V3-V4 region of 16S rRNA gene by Miseq Illumina, their reads were around 40.000 reads with the same parameter.

--> Normally, how many reads of sequences can we have after joining fastq-file of sequences V3-V4 region of 16S rRNA gene by Miseq Illumina? 

One more question is joining fastq-file by using QIIME 1.9.1 is similar to using "EA-utils". They will have same results after joining or not. I really confused here. 

Thank you very much for your kind help.

I highly appreciate your response.

Yours sincerely,

Tien.

[Greg Caporaso]

Hi Tien,

The number of reads in a MiSeq fastq file can vary widely based on the number of samples that were multiplexed on that run, the iteration of the MiSeq platform that was used (older versions of the platform produced fewer reads than newer versions), and the quality of the run. 

If the number of reads in the input (before joining paired ends) is very different from the number of reads in the output, that could be an issue with the quality of the reverse reads, or the quality of the forward and the reverse reads. My recommendation is to determine how many reads you're starting with (I don't see that mentioned here) and how many are joined after merging paired end reads. If a lot of reads are not being joined, you can either reduce the --min_overlap and/or ----perc_max_diff parameters to allow lower quality joins, or just work with the forward reads. 

fastq-join, EAUtils, and SeqPrep should all have fairly similar results. They each take a slightly different approach to joining reads, but the differences are likely to be fairly minimal. 

Hope this helps! 

Greg