Can anyone make sense of the quality scores in the qseq.txt files?

[TylerBackman]

I'm running the Illumina pipeline 1.3.2, and I noticed that the quality scores in the qseq.txt file appear to be the same as the alignment-normalized scores produced by Eland/Gerald, which are totally erroneous for reads without a good reference genome.

It seems like Gerald is going back and injecting these quality scores into the Bustard output rather than just putting them in the Gerald output. Why would they do this? Is there any way to get the quality scores in the new (Phred) format without producing the legacy prb files and converting them?

[dcjamison]

As of 1.3.2, Bustard does the qs calibration according to an internal table. You can do custom calibration from the command line, so if you want uncalibrated scores in the new format you could create probably create a custom custom calibration table that leaves the values unchanged. I do not know the format of the calibration table off the top of my head, but it is somewhere in the distribution.

[TylerBackman]

So calibration no longer depends on the reference genome?

If so, this makes a lot of sense, as it doesn't make the (often false) assumption that all failed alignments are due to sequencing error.

As of 1.3.2, Bustard does the qs calibration according to an internal table. You can do custom calibration from the command line, so if you want uncalibrated scores in the new format you could create probably create a custom custom calibration table that leaves the values unchanged. I do not know the format of the calibration table off the top of my head, but it is somewhere in the distribution.