Roche Nanopore: Accuracy
A lot has been said about Roche’s “Q39” accuracy system. I wanted to pick this apart a little more. Because… Roche Q39 isn’t the same as Illumina “80% >Q30” or Oxford Nanopore “Q30 reads”. Everyone is a little bit different.
First, let’s look at the average accuracy of existing platforms, based on my previous posts:
There’s certainly scope for arguing over these datasets, but broadly they give some sense of where the platforms sit. These numbers miss a lot of nuance, including the importance of read length, accurate quality score, systematic bias, ability to filter out errors etc. but they give some sense of where the platforms sit in relation to each other.
Where does Roche fit in the above? Roche don’t tell us. They only give the average simplex accuracy (which they’re not pushing) and the average accuracy with probable errors masked out.
Let’s explore this a little more: