> As of v1.6.1 medaka stitch can now has a [--qualities](https://github.com/nanoporetech/medaka/blob/master/medaka/medaka.py#L505) option to output a fastq file.
I was wondering how the quality scores from the network are interpreted now and if I could use the obtained quality scores for downstream analysis.
The quality scores are simply the log-transformed softmax output from the RNN. We've made little attempt to quantify how these scores match empirical residual error.
I was wondering how the quality scores from the network are interpreted now and if I could use the obtained quality scores for downstream analysis.