How are the base quality score generated? about tracy HOT 3 OPEN

@tobiasrausch

the input trace qualities are not very useful

This piqued my interest, would you mind expanding on it a bit? In my department, one of the concerns I come across as a proponent of tracy is the lack of informative quality scoring and the fact that Ns appear in our sequences at a very very low rate compared to other basecalling algorithms - combined, these attributes make my colleagues cautious.

from tracy.

The quality scoring is indeed a bit of an issue because the input trace qualities are not very useful. The assemble command simply scales a flat quality prior by the fraction of traces supporting the consensus nucleotide. For 2 input traces, it is thus indeed only 1 or 2 traces supporting the consensus nucleotide. For more input traces, you should see a range of quality values.

from tracy.

OK thanks - that's interesting. I'm using Tracy to detect errors in sequencing data, which can range from 1 trace (where I use basecall) to 4 traces (assemble).

As per your explanation, this sounds like forming a consensus between 2 traces for a given nucleotide doesn't consider the quality of the base call, and rather just looks at the fraction of traces involved in generating the consensus.

Below summarises my understanding for 4 different base quality configurations for the assembly of 2 trace files - is this accurate? To my mind, the 3rd and 4th scenarios should have lower quality values than the 1st.

Part of the problem for me is that I want to have some estimate of the per-base quality score, so that I can confidently calculate the per-base error rate. In practice, this is hard using tracy because the quality scores change depending on how many trace files I use, and don't seem too comparable between a 2-trace assembly and a 4-trace assembly.

Is there a workaround?

from tracy.