non-matching barcodes

[PFS]

Hi,

I have some RNASEQ data from Illumina 1.4+ pipeline.
The barcodes of a multiplex run are appended to the header line of the fastq file. I supposedly have two samples in there, so I am expecting two barcodes, however, I see also many other barcodes different from what I expected. The counts of these "spurious" barcodes is very low, so the majority of the sequences are associated to the expected barcodes.

When splitting the fastq file according to the barcode, how are the sequences with non-matching barcode handled? Do I allow 1 mismach? Do I keep exect matching criteria?

Thanks in advance.

[HESmith]

The standard Illumina barcodes are designed to tolerate a single mismatch and still be assigned properly. The spurious barcodes you observe are typically associated with clusters that yield low-quality reads.

[PFS]

Thank you, HESmith!

The barcodes are in the header so I am assuming I cannot use the utility in fastx_toolkit.

Does it make sense to list all possible barcodes that would be acceptable (i.e. those with one mismatch only, in any position) and then split the file according to this file?

For example:
Barcode1: ACGT
Acceptable1: ?CGT, A?GT, AC?T, ACG?

Barcode2:AGGT
Acceptable2: ?GGT, A?GT, AG?T, AGG?

Then the two output files will have the reads that match the acceptable barcode sets, while all other reads will be discarded?

Thanks!

[HESmith]

You can parse the reads this way (although your example is not ideal b/c A?GT is a match for both barcodes), or run the demultiplex.pl script (part of Illumina's CASAVA package) with the "--mismatches 1" flag.

[PFS]

or run the demultiplex.pl script (part of Illumina's CASAVA package) with the "--mismatches 1" flag.

Thanks! That's what I needed!