Install JKBC

1. Cd into the repository folder: cd basecaller-p10
2. Create a conda environment from the conda_env.yml file in the repo:
   • Create a new conda environment using the conda_env.yml file: conda env create -f conda_env.yml (this will take a while)
3. Activate the conda environment: conda activate jkbc
4. Install the JKBC library locally using pip: $CONDA_PREFIX/bin/pip install -e jkbc

Making predictions

1. Activate the conda environment if not active: conda activate jkbc
2. cd into basecaller: cd nbs/basecaller
3. Run the prediction script python predict.py <id> <data_set> <name_of_run>

• Any id from https://app.wandb.ai/jkbc/jk-basecalling-v2 can be used, however the models presented in the report are:
  • JKBC-1: 2eiadj4y
  • JKBC-2: 1ywu3vo9
  • JKBC-3: 2d84exku
  • JKBC-4: j6f2sn3v
  • JKBC-5: 1c2vr2my
• A small test set is include in nbs/basecaller/test-data/
• To predict using JKBC-5 use the command: python predict.py 1c2vr2my test-data
• This creates the folder 1c2vr2my-test-data/ containing reference.fasta and predictions.fasta.