Demultiplexing softwares

[ClemBuntu]

Hello everyone,

I used to demultiplexe using bcl2fastq.pl provied by CASAVA.
I want to try new tools because I got some trouble at the demultiplexing step.
So I checked on google and I found several softs like TagGD, Flexbar, Sabre, deML etc. but all of them take a FASTQ file in input... And the only way I can have a FASTQ file is demultiplexing using bcl2fastq.pl plus this script removes the index sequence of all reads...
So how can I demultiplexe something which is already demultiplexed and where the index sequence are removed ?

I think the only thing I can do is to give an empty SampleSheet to CASAVA, then I suppose it will put all reads together as undetermined reads, but maybe there is a cleaner way to do ?

[GenoMax]

Can you specify what kind of "trouble" you ran into with bcl2fastq?

Please keep in mind that if there was a problem with the index reads (multiple N's in the sequence) then no demultiplexing tool is going to help you. This run may have to be repeated.

[ClemBuntu]

Well I got more undetermined reads that I excepted.

I'm sure the main problem comes from the run but anyway how can I do if I want to compare these toolts ?

[westerman]

Well I got more undetermined reads that I excepted.

I'm sure the main problem comes from the run but anyway how can I do if I want to compare these toolts ?

I suspect (not having used the tools you mentioned) that they are focused on demultiplexing user-generated barcodes instead of using the Illumina indexes and thus you will be disappointed in them. As you said you could force everything into undetermined reads but that won't give you the Illumina indexes.

If you have more undetermined reads than expected then, like GenoMax, I think that you have problems with your indexing reads. You can alleviate some of the problems via the 'use-bases-mask' and 'num_of_mismatches' parameters to Casava.

[GenoMax]

Well I got more undetermined reads that I excepted.

I'm sure the main problem comes from the run but anyway how can I do if I want to compare these toolts ?

What fraction of the reads are in the undetermined file?

Are you recovering all the samples you expect from the run? A common problem is specifying an incorrect barcode for a sample which sends the reads for that sample(s) to the undetermined file.

[westerman]

GenoMax has, as usual, great suggestions.

His "common problem" of specifying an incorrect barcode (which is indeed something we do often enough) should be easy to spot because one of your samples will have close to zero reads.

The harder problem is if many of the index reads overall have poor quality bases in them. If the problem is base-specific (e.g., occurs only at the 2nd index base) then Casava's 'use-bases-mask' parameter can be used to ignore the specific base. Otherwise 'num_of_mismatches' needs to be used. Sometimes I use both if the reads are really poor.

[GenoMax]

Use the following command to find out what tags are represented in your "undetermined" file. You will generally see a great variation present. Thing to look for is if there are some tags that are way over-represented. Sometime people make mistakes in making libraries/with pooling and you don't quite have the result that you expect from a pool.

Replace the ? with specific lane number.

$ zcat lane?_Undetermined_L00?_R1_001.fastq.gz | grep @HWI |cut -d: -f10 | sort | uniq -c | sort -r -n -k1

[ClemBuntu]

The fraction of undetermined reads is not too high, so don't worry about that guys but thanks anyway....


My question was about demultiplexing tools that take FastQ files as input... I'm just wondering why they exist since Illumina software like bcl2fastq demultiplexe and creats FastQ file.... It seem nobody here use these tools that's very strange !

As westerman said maybe theses tools are only useful when you have a custom barecode...

[Michael.Ante]

Long before the index reads read-out, barcodes were inserted into the reads' sequence (inline barcoding). You had to demultiplex the Fastq file and split it into the different samples according to the first n bases. Therefore, various tools were implemented.
Now in times of the index reads, these tools are more or less outdated.
I hope that answers your question.
Cheers,
Michael

[ClemBuntu]

Oh I see, now that makes sense thanks !