Badread crashes with qscore model unset

[nrizzo]

Hello,

The instructions to run the GraphChainer experiments specify to download the latest Badread version (git clone git clone https://github.com/rrwick/Badread.git), but doing so results in an error when running generate_sim_reads.py:

$ python3 generate_sim_reads.py -g graph.gfa -fq sim_reads.fastq -p sim_reads_path.nodes -fa sim_reads_path.fasta
[...]
Loading error model from [...]/GraphChainer-scripts/Badread/badread/qscore_models/pacbio2016.gz
  done: loaded error distributions for 16384 7-mers

Loading qscore model from [...]/GraphChainer-scripts/Badread/badread/qscore_models/nanopore2023.gz
  X=DXXX==  Error: [...]/GraphChainer-scripts/Badread/badread/qscore_models/nanopore2023.gz does not seem to be a valid qscore model file

This is because the error model pacbio2016 is picked explicitly, whereas the default qscore model — now nanopore2023 — is loaded, and the two seem incompatible. For reproducibility of the GraphChainer tests, I suggest fixing the Badread version with a submodule (and adding the explicit qscore model anyway). After this, a GraphChainer tag could be added so that the pipeline can be updated with newer Badread version/new features.

Best,
Nicola