NeoantigenEditing

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:

1. Align each patient's neoantigens to IEDB

   python align_neoantigens_to_IEDB.py

2. Compute neoantigen qualities and fitness of all clones

   python compute_fitness.py

3. Predict clone frequencies in recurrent tumors:

   python predictions_clones.py

4. 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:

• Zachary Sethna

• Marta Luksza

• Ben Greenbaum

• Vinod Balachandran