The abundance of alpha-fetoprotein (AFP), a natural protein produced by the fetal yolk sac during pregnancy, correlates with lower incidence of estrogen receptor positive (ER+) breast cancer. The pharmacophore region of AFP has been narrowed down to a four amino acid (AA) region in the third domain of the 591 AA peptide. Our computational study focuses on this 4-mer segment consisting of the amino acids Threonine-Proline-Valine-Asparagine (TPVN). We have run replica exchange molecular dynamics (REMD) simulations and used 120 configurational snapshots from the total trajectory as starting configurations for quantum chemical calculations. We optimized structures using semiempirical (PM3, PM6, PM6-D2, PM6-H2, PM6-DH+, PM6-DH2) and density functional methods (TPSS, PBE0, M06-2X). By comparing the accuracy of these methods against RI-MP2 benchmarks, we devised a protocol for calculating the lowest energy conformers of these peptides accurately and efficiently. This protocol screens out high-energy conformers using lower levels of theory, and outlines a general method for predicting small peptide structures.