Comparison of PredictedpKaValues for Some Amino-Acids, Dipeptides and Tripeptides, Using COSMO-RS, ChemAxon and ACD/Labs Methods

Abstract

Liquid-phase <i>pK<i/><sub><i>a<i/><sub/> values play a ke<sup><i>y<i/><sup/> role in food science. Chemical properties of molecules depend largely on whether they are ionized or not. Most organic molecules are capable of gaining and/or losing a proton in aqueous solutions. Proton transfer most. frequently occurs between water and any ionizable atom of the organic molecule. The molecule’s response to profanation or deprotonation depends significantly on the site that was disturbed by proton transfer. Partial charge distribution in the molecule also varies with protonation of the acidlbase active sites. Then it can he used to determine the <i>pK<i/><sub><i>a<i/><sub/> of a molecule. First, we use the COSMO-RS method, a combination of the quantum chemical dielectric continuum solvation model COSMO with a statistical thermodynamics treatment fin- more Realistic Solvation (RS) simulations, for the direct prediction of <i>pK<i/><sub><i>a<i/><sub/> constants of about 50 molecules (amino-acids, dipeptides and tripeptides). Then, we compare our results with experimental data and the <i>pK<i/><sub><i>a<i/><sub/> values predicted using two other methods. We used respectively the ChemAxon method using a program based on the calculation of partial charge of atoms in the molecule and the ACD/Labs method that enables to calculate single <i>pK<i/><sub><i>a<i/><sub/> values. for all possible dissociation centers when the rest of the molecule is considered neutral, using an internal database containing chemical structures and their experimental <i>pK<i/><sub>a<sub/> values. The averaged Root Mean Square Error (RMSE) of the predicted <i>pK<i/><sub><i>a<i/><sub/> values for each method compared to experimental results were respectively 0.596 for COSMO-RS, 0.445 for ChemAxon and 0.490 for ACD/Labs. While ACDILabs and ChemAxon are parameterized using a large set ofexperimental data (including several of the studied molecules), the COSMO- RS method was used in a fully predictive way. Regarding these results, COSMO-RS appears as a promising method to predict the <i>pK<i/><sub><i>a<i/><sub/> values of molecules of interest in food science with scarce available <i>pK<i/><sub><i>a<i/><sub/> values such as peptides. The final goal of this study is to use the <i>pK<i/><sub><i>a<i/><sub/> values in a predictive thermodynamics model for products of interest in food industry. For this purpose, the effects of several factors (like conformations set treatment in COSMO-RS calculations, ionic strength effect) that can affect the comparison between observed and predicted <i>pK<i/><sub><i>a<i/><sub/> data are discussed.