Organ transplantation is the sole treatment for end-stage organ failure, with 10,810 patients on the waiting list and only 5,493 transplants conducted in France in 2022. The primary cause of transplant loss is the recipient’s alloimmune response, particularly the humoral response producing donor-specific antibodies (DSA) against Human Leukocyte Antigen (HLA) mismatches. The complexity of the HLA system, with its vast genetic polymorphism, poses challenges for defining compatibility between donor and recipient, which currently relies on serological classifications at antigen level.

This thesis addresses the challenge of defining HLA antigenicity through 3D molecular data and data science. High-quality 3D structures and molecular dynamics (MD) simulations of 207 HLA antigens were generated and analyzed to understand structural differences influencing antigenicity. Complementarily, 3D shape descriptors for comparison of dynamic protein surfaces (i.e. Zernike descriptors) were explored. Additionally an HLA Epitope Predictor (HLA-EpiCheck)  was trained on static and dynamic structural features to predict HLA antigenicity. The predictor demonstrated a high performance in identifying eplets, showing strong agreement with experimental data. Finally, HLA-3D-Diff was developed to visualize and compare 3D HLA structures, aiding in the understanding of complex immunization patterns.

Keywords: Organ transplantation, molecular dynamics, HLA antigenicity, data science.