Neoantigen quality predicts immunoediting in survivors of pancreatic cancer, Nature 2022
Code for computing neoantigen qualities and for performing clone composition predictions.
The input data are the following:
data/Patient_data - folder with phylogenies for each of the patients. Top 5 scoring trees are provided for each patient. Tree clones are annotated with mutations, predicted neoantigens and clone frequencies.
data/epitope_distance_model_parameters.json - cross-reactivity metric
data/fitness_weights.txt - optimized fitness model weights for each recurrent tumor.
data/iedb.fasta - IEDB epitopes used for the analysis in the paper (downloaded from the IEDB on January 2022)
To run the code:
- Align each patient's neoantigens to IEDB
python align_neoantigens_to_IEDB.py
- Compute neoantigen qualities and fitness of all clones
python compute_fitness.py
- Predict clone frequencies in recurrent tumors:
python predictions_clones.py
- Compute log-likelihood scores - comparison between the fitness model and the model of neutral evolution of tumors.
python predictions_aggregated_loglikelihood_scores.py
For any questions please contact: