Structural Immunoinformatics
3pHLA-score
3pHLA-score improves structure-based peptide-HLA binding affinity prediction
Large dataset of more than 70,000 pHLA modeled structures
We used our tool APEGen to model over 70,000 structures and map them to binding affinity data from the IEDB. We used this dataset to train regression models for binding affinity prediction.
Unique per-peptide-position featurization
3pHLA-score develops and uses a per-peptide-position featurization decoupling standard terms from Rosetta ref2015 score into separate structural terms for each peptide position.
Making 3pHLA-score accessible
We provide a python package for scoring pHLA structures and show use-cases within Google Colab Notebooks.
