PacBio RS quality score output definition (FASTQ)

[vamiyn]

I'm simulating reads that mimic PacBio's data (long reads, about 1k-20k bases). Other things are pretty well in order, but I'd like to know what kind of quality values there usually are and what their distribution and meaning is. I'd be very thankful if someone showed me to the right direction or explained at least some part of this!

- which characters are used? (as there seems to be many kinds of variations of FASTQ... I just can't seem to find anything about PacBio's FASTQ directly)
- is there a bigger probability for an indel (or a substitution) if the quality score is bad?
- how often do you generally see certain quality scores, where goes the line between "probably an error" and "most likely fine" ?
- how likely is it to have a bad quality score if the read itself is errorless?

Thank you in advance!

[mchaisso]

PacBio produces reads that have different quality values for deletion, insertion, and substitution, so the read are in HDF5 rather than FASTQ.

There is a read simluator called alchemy that is packaged with blasr that simulates HDF5 in a large number of ways using an empirical error model.

[GenoMax]

You can find useful PacBio software here (including blasr mentioned by mchaisso): https://github.com/PacificBiosciences

[vamiyn]

Okay, I see... I have to use FASTQ for this, so unfortunately the HDF5 reads alone don't help that much. (PacBio really doesn't produce FASTQ directly?) Thank you, though, I'll definitely check the link anyway!

But okay, now I'd need to know what happens if the reads are converted to from HDF5 to FASTQ format afterwards, what will happen to which quality value or is there a standard for that? Seems like at least some information would be lost then? Or does FASTQ's quality string have enough information (compared to HDF5) that it's worth mimicking? (Does blasr use the quality string in FASTQ or only in the HDF5?) Plus the questions in my first post, still, if FASTQ has enough info.