Demultiplexing Casava 1.8 reads

[Metz]

Hi all,

My most recent data set consists of a set of fastq files that have already been demultiplexed by Casava 1.8. However, my sequencing core told me that this only allows for a single bp mismatch. Because my barcodes are rather long many reads end up in the unknown indices. However, I designed the barcodes to allow for 5 errors between them and I would now like to utilize this. I can easily look at the fastq files and see the index in the read name header, but I can not find a program that will allow me to demultiplex based on this format. My core doesn't keep the original reads, so I am stuck having to demultiplex reads that have already been demultiplexed by Casava. If anyone knows of a program that can accomplish this that would be great.

[Frank D]

Hi Metz,
Try the fastx_barcode_splitter.pl from the FASTX toolkit:
http://hannonlab.cshl.edu/fastx_toolkit/
You can specify the number of mismatches and for a one off analysis it is reasonably quick.

[Metz]

Thanks for the response Frank, but I've already tried fastx. Unless I'm mistaken, it requires that the barcode be part of the read, not in the read name. I've thought about just writing a script to reintroduce the barcode back into the read, but that also requires adding in 'mock' quality scores for those bases and other changes to the read name. My perl is several years out of use, and I'm trying to prevent reinventing the wheel.

[GenoMax]

I assume you are looking to parse reads from the "Undetermined" reads file which would have the reads with more than 1 mismatch.

Rather than re-introducing the tags back in the reads it would be more efficient to enumerate all "tags" that are in your "undetermined tags" file and decide the ones you want keep/extract.

[Metz]

Do you know of a program that can do something along those lines?

[HESmith]

The following should work for CASAVA v1.8 fastq files (the grep expression may need to be modified to match your machine name):

zcat name.of.fastq.gz | grep ‘^@HWI’ | cut -d : -f 10 | sort | uniq -c | sort -nr > indices.txt

[Metz]

Thanks for the code. It works for giving me a list of barcodes and their counts. I can definitively tell which barcode the most abundant ones belong too. However, I'm not sure how to proceed with this. I just don't understand enough perl to move forward quickly. However, if there isn't another solution, I guess that is the way to go. I just can't believe that nobody else has had this problem in the last year.