PacBio CCS vs subreads explained ?

[curious.genome]

Hi guys,

I've been trying to understand PacBio data for a while now, I think that it's the way forward for our lab. I'm still having trouble understanding CCS though - my biology is quite shaky.


1. Could someone explain to a computer scientist - how does PacBio get its CCS from the subreads or long reads ?

2. Our collaborators with the PacBio data gave us a set of files - with subreads.fastq, CCS.fastq and long_reads.fastq. When I ran a FASTQC report for subreads, I got base quality scores of around 10-15, whereas with CCS I got a larger variation for quality scores with values between 30-40. Is this expected ?

3. Which of these files should I use to perform assembly ? I'm guessing it's CCS.fastq


Sincerely,
Confused-in-pacbio-land

[mcnelson.phd]

I'm just really beginning to dabble in PacBio myself, so I'm not an expect but this is my understanding of things:

Subreads are the individual sequence reads that are determined in real time from a template on the sequencer. These would correspond to the stretch of DNA between the adapter sequence.

CCS reads are the result of doing a consensus base calling from subreads that are all from the same template. If the template was short enough, then the polymerase will loop entirely from the beginning back to the beginning, sequencing the adapter before it starts on the template again. In this case, the same template piece of DNA is sequenced more than once, which means that the sequence data generated from each pass can be used to determine one single consensus sequence with higher base quality than the raw subreads. Thus, the quality distribution that you see is correct between the subreads and the CCS reads

The long reads are reads that did not get through the entire template as the adapter sequence was found. These reads would be unique compared to the CCS reads and would have similar low Q-scores as the subreads.

For assembly, the question of which reads you use would depend on how you assemble. Using the HGAP assembler, you essentially use all of the read files. If you're planning on assembling with another program, then you need to carefully think about how you'll use the low quality long reads in conjunction with the shorter but higher quality CCS reads.

[GenoMax]

Lots of PacBio training resources are available on this site: https://github.com/PacificBioscience...-Training/wiki

Some of your questions will be addressed by this set of slides: https://speakerdeck.com/pacbio/speci...equencing-data (slides 24-28).

[curious.genome]

Thanks for taking the time to explain that, mcnelson, it definitely helps ! I can now proceed to my next step - investigating how to work with the HGAP assembler.

GenoMax thanks for the links to the resources (esp. pointers to slides 24-28).