This is a test version of Biostars. For the public version, visit https://www.biostars.org.
MinION nanopore sequencing quality

the MinION quality scores do not follow Phred expected error rates, according to a paper that is Assessing the performance of the Oxford Nanopore Technologies MinION .Each base has a quality score, I want to know how to compute the MinION quality scores .

sequencing

Your question is unclear, please provide more information.

I have updated this question,thanks for any suggestion.

Quality scores corresponding to nucleotides are calculated by the software used for basecalling. Dependent on when the data was generated this is either a Hidden Markov Model (HMM) or a Recurrent Neural Network (RNN). You can't compute those basecall quality scores yourself, and I also don't understand why you would want to.

Since you are still vague in your question concerning which paper this is from I cannot help you further with that.

My question might not be clearly stated. I know I can't compute those basecall quality scores.I just want to know the detailed algorithm for quality scores.Thanks

Then try to be complete from the beginning when you ask your question.

I don't think the exact algorithms are disclosed by ONT.

That's why I can't find the related information.Thank you very much.

So are those quality scores different than what Illumina ould have? like the Q-score?

The idea is the same: likelihood of having that nucleotide at that position. But as the chemistry and signal is fundamentally different the calculation of phred scores will be different as well. But they mean the same.

Good point. I didn't see the other thread (where you had essentially the same suggestion).

I assumed he was working with his own reads.

1 answer

You can use an open source basecaller (such as Nanocall) if you want to know the exact algorithm. The results are supposed to be comparable to the official ONT basecaller.

Nanocall has a fair amount of background info on their website: https://github.com/mateidavid/nanocall

This is interesting to see, thanks.

However, it seems that since the release of Nanocall, ONT have open-sourced Metrichor, and the quality of nanocall doesn't hold up on the latest chemistry.

From the nanocall github:

If you want to use Nanocall on R9 data, Nanocall does support it directly, but its accuracy is significantly lower than that of Metrichor (unlike the case of R7.3, where the two had similar accuracy). The reason for the discrepancy is that Metrichor on R9 uses a more elaborate RNN-based approach, compared to the simple HMM-based one in Nanocall.

The Metrichor source code was opened (under a development license).

Which is pretty damn restrictive...

Oh. I didn't see that. Good to know.

Log in to answer this question.