Problem of demultiplexing 4 barcoded E.coli genomes in one Sequel SMRT cell

[weigrc]

Dear all,

As captioned, this sequencing included 4 barcoded libraries prepared by the following protocol, and barcode lbc01, lbc02, lbc03 and lbc04 were chosen, respectively. The library size is around 7.5kb (the longest subread is 5259 bp), and the SMRT Link we use is Version 4.

http://www.pacb.com/wp-content/uploa...Sequencing.pdf


Refer to attached, demultiplexing the data with full set of 96 barcodes, the problem is that most of the sequencing reads are sorted as other 92 no-existing barcodes, but the 4 barcodes we used!!!!

I think the size as 5,259 bp is too long for PacBio demultiplex.


Anyone has experience on sequencing barcoded libraries?? suggestions?


Thanks,
Wei

[GenoMax]

Have you contacted PacBio tech support directly about this?

While it is common to see some sequences getting assigned to indexes you did not use, you should obviously get the correct results, if the experiment worked as designed.

[jharting]

What barcode set did you use for barcode scoring? The v4.0 software is orientation specific. If you used barcodes in the wrong orientation, they will not be binned correctly.

[weigrc]

We were using PacBio SMRTbell Barcoded Adapter Complete Prep Kit-96.

And we did contact our tech support to follow up this issue since this Tuesday with the full information they needed, but the answer I got so far is only PacBio does not recommend to do barcoded WGS.

I understand some reads are possibly mis-assigned to undersigned barcodes, but I worry about the possibility is much bigger than my expectation, I was told 1% or more may be normal for PacBio sequencing. Is it true?

Anyone has experience on barcoded WGS using Sequel system?


Thanks,
Wei

[jharting]

We have been doing multiplexed microbial assembly on Sequel with success. Barcoded adapters are for general use -- basically a smrtbell adapter with extra sequence, so there's no reason why they cannot be used for your purpose.

You should be using the barcodes labeled something like 'pacbio_barcode_adapters_96' for scoring them. That set is in the correct orientation. If you have used barcoded adapters previously with smrt analysis version 2.3 or earlier, you may have used a barcode file with the sequences in the other orientation without issue. That software was not orientation-specific, and so it did not matter. The barcoding tool was re-written for v3+, and it is now orientation-specific.

[weigrc]

Thank you, jharting.

Do you know what algorithm SMRT Link is using for demultiplexing? I mean it requests CCS read or full-pass subread? Is it possible to access such setting?

The barcoded WGS library I've tested has 7.5 kb peak size with smear size range 3k-10k, and the polymerase read length shown in the attached. Although the official protocol I followed is for 10 kb library, I guess the 7.5 kb library is still no short enough for a good demultiplexing result. May I know what library size you have been targeting for barcoded microbial WGS?

Regards,
Wei

[jharting]

Hi Wei,

We typically target libraries of ~10-12kb for multiplexed assembly, although there is work to optimize this, including a size selection step to remove short reads and pre-extension runs where the laser is turned off for ~120min to let the polymerase get near the end of the insert before sequencing begins. The trade-off in multiplexed assembly on PB platforms is between length and barcode yield. You want long reads to generate nice, closed assemblies, but too long and the polymerase is less likely to reach a smrtbell adapter/barcode before dying, in which case you may get a nice long read but no barcode for binning!

The underlying tool in smrtlink for barcoding is 'bam2bam'. You can find some information about the tool in the smrt tools reference guide:
http://www.pacb.com/wp-content/uploa...ide-v4.0.0.pdf

This tools works only on the raw subreads. To label a zmw (sequencing hole) with a barcode, at least one smrtbell adapter must be included in the hqregion of the polymerase read for the zmw.

Here also is a link to a "third-party" demultiplexing tool that you can use, if for nothing else than to look at the demultiplexing report, which can be useful for diagnosing your dataset.

https://github.com/armintoepfer/uhu

Although this is technically "third-party" because is exists outside of pacbio an is not supported, its written by the same developer who wrote bam2bam. I find it useful and fast.

[weigrc]

Hi Jharting,

Thank you for these information, and I am reading the documents, they seem useful. thanks!

[CarambaKaracho]

Hi weigrc and jharting, just wanted to thank you for the concise information regarding pacbio multiplexing. Your excellent questions and answers helped me a lot to understand what I'm getting from a sequencing provider.

[jharting]

Here is a beta version of the next generation barcoding tool for pacbio if you are interested in trying. It has a lot of useful options and a nicer user interface!

https://github.com/PacificBiosciences/barcoding

[dho]

I know this is a while after the last reply, but I struggled for much of the summer with a similar issue. Basically, if the barcode is not immediately adjacent to the SMRTBell adapter, the demultiplexer in SMRTLink behaves inconsistently. There are two solutions:

1) Make CCS files and demultiplex manually by looking for the sequences of your barcodes in the CCS files.

2) You can use the new PacBio Lima demultiplexer (the one mentioned by jharting) with the -W flag to set a larger window for barcode detection. For example, in my data, there is about 30bp of sequence between the adapter and 15bp barcode. Setting -W 50 allows barcodes to be detected correctly.

--dave