The implementation of machine learning techniques in the field of immunoinformatics for epitope prediction underpins the complexity of these processes. The problem is further complicated by the enormous diversity of the HLA alleles in humans and the resulting variety of epitopes that can possibly be identified by the immune systems of different patients. In order to produce deimmunized biotherapeutic drugs effective in all patients, all these different alleles would need to be considered. In this context, an interesting option may be the development of personalized deimmunized biotherapeutics, taking into consideration only individual HLA alleles. These personalized deimmunized drugs, however, would generate high costs for patients and/or healthcare systems (even though the use of computationallydriven methods can reduce costs for drug development significantly) and likely face regulatory obstacles. In general, computationally-driven methods for epitope prediction depend on the quality of the input data that are provided, the degrees of freedom allowed, and the algorithms used. Nevertheless, computational methods have already been reported to be more accurate than experimental approaches in epitope prediction in some cases, e.g., for the development of peptide-based vaccines [3-5].