For exposition purposes, the entire pipeline has been laid out in a series of iPython notebooks:
| Notebook | Contents |
|---|---|
| 1-data-prep.ipynb | Data transformation and filtering |
| 2-data-split-tcr.ipynb | Generating negative training examples by randomly mispairing TCRs and epitopes and train / test splitting of the data (ensuring mutual exclusivity in TCRs) |
| 3-tcr2vec.ipynb | Training a tcr2vec embedding model for TCRs using gensim |
| 4-model-tcr.ipynb | A Tensorflow implementation of the TCR-pMHC binding prediction model |