Human leukocyte antigen (HLA) is an important molecule family in the field of human immunity, which recognizes foreign threats and triggers immune responses by presenting peptides to T cells. In recent years, the synthesis of tumor vaccines to induce specific immune responses has become the forefront of cancer treatment. Computationally modeling the binding patterns between peptide and HLA can greatly accelerate the development of tumor vaccines. However, most of the prediction methods performance is very limited and they cannot fully take advantage of the analysis of existing biological knowledge as the basis of modeling. In this paper, we propose TripHLApan, a novel pan-specific prediction model, for HLA molecular peptide binding prediction. TripHLApan exhibits powerful prediction ability by integrating triple coding matrix, BiGRU + Attention models, and transfer learning strategy. The comprehensive evaluations demonstrate the effectiveness of TripHLApan in predicting HLA-I and HLA-II peptide binding in different test environments. The predictive power of HLA-I is further demonstrated in the latest data set. In addition, we show that TripHLApan has strong binding reconstitution ability in the samples of a melanoma patient. In conclusion, TripHLApan is a powerful tool for predicting the binding of HLA-I and HLA-II molecular peptides for the synthesis of tumor vaccines.