In the context of organ transplantation, recipient’s antibodies against donor-specific HLA antigens are the main reason for transplant loss. The prediction of B-cell (antibody) epitopes on HLA antigens is therefore a key challenge on the way to improving the matching step between donor and recipient from a structural point of view. Here, we present HLA-EpiCheck, a B-cell epitope prediction tool that relies on an unprecedented dataset of short Molecular Dynamics (MD) simulations of 207 HLA antigens. We use hydrophobic properties, electrostatic charges, flexibility and solvent accessibility as descriptors calculated on patches sampled from MD trajectories. Then, we train an Extremely Randomized Trees machine learning model. This model outperforms the state-of-the-art DiscoTope 3.0 tool for B-cell epitope prediction on HLA antigens.