FURI | Spring 2023

Impact of Position Information on TCR-epitope Binding Affinity Prediction Model and Its Biological Significance

The binding of T cell receptor (TCR) and epitope plays an important role in the immune system. Machine learning models have been developed to predict the binding affinity of TCR-epitope pairs. One of the existing models using the Transformer architecture has shown a great capability for this task. However, the position information of TCRs and epitopes is missing during the computation due to the lack of a positional embedding layer. Researchers aim to add the positional encoder to the current model to see its effect on the prediction, and to discover the biological significance of the position information to the binding affinity.