PASC analysis

Dependencies

The mamba/conda environment that contains all dependencies for this project can be created with

conda env create -f env.yml
conda activate pasc
pip install -e .

or

mamba env create -f env.yml
mamba activate pasc
pip install -e .

Pipeline

Processing the data

python get_training_data.py --infile ADAPTIVE_ANNOTATIONS.tsv \
                            --igor_wd /PATH/TO/IGOR/WORKING/DIRECTORY \
                            --outdir /PATH/TO/SAVE/OUTPUT \
                            --infer_unproductive_alignments

will process the Adaptive annotations, producing files giving the unique allelic-level VDJ recombinations for unproductive and productive annotations as well as csvs to be used as input for SONIA/SoNNia models.

See the container_files directory for submitting jobs using a containerized minimamba (similarly for miniconda).

To produce IGoR training data, unique allelic-level VDJ recombinations are obtained across timepoints using

python combine_igor_data.py --indir /PATH/TO/PROCESSED/UNPRODUCTIVE/FILES

Aggregating data

To obtain training data for cohort-level models, use

python aggregate_training_data.py --indir /PATH/TO/PROCESSED/FILES \
                                  --outdir /PATH/TO/SAVE/OUTPUT

aggregate_training_data.py has options for subsampling to get training data as well as changing the of pasc. Use

python aggregate_training_data.py -h

to see all the options.

IGoR

To produce IGoR models, run

bash container_files/submit_igor_jobs.sh

if using an Apptainer container that has minimamba with the environment created. Otherwise use

bash run_igor.sh WORKING_DIRECTORY SEQS_FOR_IGOR_FILE BATCHNAME

SONIA/soNNia

Population Dynamics Analysis with NoisET

To perform population dynamics analysis with NoisET on repertoire data, first download noiset_analysis.py and noiset_loop.sh. To run analysis on repertoire data, run the following

bash noiset_loop.sh file1 file2

where the first file should be the first timepoint and the second file should be the second timepoint. Update noiset_analysis.py as necessary so that files save to an appropriate directory.

To run population dynamics analysis on all INCOV repertoire data, download submit_noiset_jobs.sh and run the following:

bash submit_noiset_jobs.sh