Subreads alignment alternatives

[danova]

Hi,
Just wondering if anybody has tried to align subreads and build the CCS using alternative alignements programs, others than Blasr.

I´m thinking of bwa.
The problem is that if you are working with amplicon 16S data and don´t have a reference how would that work. So far bwa mem requires a reference database.


So anybody used alternative alignement programs to build consensus without a reference database ??

[rhall]

Building consensus for a CCS read does not use Blasr. The consensus generation is based on a detailed model of the PacBio error profile. What would be the aim of using an alternate consensus caller?

[danova]

I´d like to aligne the subreads by other means and compare it to the obtained CSS from the pacbio and evaluate the error rate.
Don´t know how to actually get the error rate as well as number of mismatches, insert and deletions from the bax.h5 files ??

[rhall]

What are you evaluating the error rate of? Without a reference it's impossible to calculate the error rate of the CCS reads.

[danova]

Right I mixed a coupe of things here. Let me explain better
1 / I have sequenced a community mock and generated the CSS. From the CSS i can define the error rate.
2 / Now, I would like to generate CCS by other means and see how the pacbio performs when compared to other alignement programs. I don´t know how to generate the consensus using other soft.

If i don´t compare the pacbio alignement tool to other external soft I don´t know if the CSS generation can be improved.

[rhall]

I don't know of any other program that would be able to generate a consensus even close to the quality of the PacBio CCS algorithm due to the the use of the error profile and complex quality values. The only workflow that may be worth testing is quiver (PacBio's multimolecule consensus algorithm) for that you would have to extract the subreads from a single molecule and run quiver on the subreads, but the CCS algorithm is based on the same concepts as quiver so it is unlikely that the results would be significantly better, as CCS has been optimized for data from a single molecule.
Note PacBio has a new CCS algorithm currently in beta, that generates better results than the current production CCS https://github.com/PacificBiosciences/pbccs although it is not currently usable without direction from PacBio.

[danova]

Thanks,
It looks to me that i´d better wait for Pacbio to work on that and see how it improves over the time. Thanks for the update,
david