MHC class I molecules report on the health of the cells through the presentation of cytosolic peptides to the cytotoxic T-cells. MHC-I proteins are highly polymorphic with over 10,000 allotypes (termed HLAs, Human Leucocyte Antigens), each displaying a distinct repertoire of peptides. Therefore, a person’s HLA genes, and thus the collection of peptides that can be presented, determine one’s ability to respond to certain pathogens or malignancies and is an important source of variation across the human population. In this work, we report on the creation and analysis of HLA3DB, an annotated database of high-resolution pHLA structures. We find that variations between peptide backbone conformations can be captured primarily by deviations in the peptide backbone dihedral angles from position 4 to 7. A complex interplay between this region of the peptide and the HLA amino acids leads to similar peptide backbone conformations from seemingly unrelated allotypes and peptide sequences. We plan to utilize this database to conduct homology modeling and predict novel pHLA structures from a given amino acid sequence.