Theoretical and quantitative prediction of pK_a values at low computational cost is a current challenge in computational chemistry. We report that the isodesmic reaction scheme provides semi-quantitative predictions (i.e. mean absolute errors of 0.5–1.0 pK_a unit) for the pK_a1 (α-carboxyl), pK_a2 (α-amino) and pK_a3 (sidechain groups) of a broad set of amino acids and peptides. This method fills the gaps of thermodynamic cycles for the computational pK_a calculation of molecules that are unstable in the gas phase or undergo proton transfer reactions or large conformational changes from solution to the gas phase. We also report the key criteria to choose a reference species to make accurate predictions. This method is computationally inexpensive and makes use of standard density functional theory (DFT) and continuum solvent models. It is also conceptually simple and easy to use for researchers not specialized in theoretical chemistry methods.