QSPR models predicting pK_a from atomic charges

T. Bouchal, R. Svobodová Vařeková, S. Geidl, M. Kudera,

O. Skřehota, C. M. Ionescu, and J. Koča

National Centre for Biomolecular Research and CEITEC - Central European Institute of Technology, Masaryk University, Kamenice 5, 625 00 Brno-Bohunice, Czech Republic
357534@mail.muni.cz

The acid dissociation constant pK_a is one of the fundamental properties of organic molecules. The pK_a values are of interest in chemical, biological, environmental and pharmaceutical research. One critical feature is that pK_a values help discover whether molecules can be used as drugs. Moreover, pK_a values are essential for ADME profiling, give insight into interactions of drugs with a receptor, etc..

Several methods for pK_a calculation have been developed [1], but prediction of pK_a values remains a challenge. A very promising approach for pK_a prediction is the usage of QSPR (Quantitative Structure Property Relationship) models which employ partial atomic charges [2,3,4]. The accuracy of pK_a prediction by these QSPR models is influenced by many factors. Very important factors are proper selection of descriptors, usage of relevant charge schemes, influence of a molecular structure, etc..

In our work, we analyzed the key factors which influence the quality of QSPR models for pK_a prediction. We focused on three types of molecules – phenols, benzoic acids and anilines. For each type of molecules, we studied the influence of the charge calculation scheme on the quality of QSPR models. Specifically, we tested combinations of three QM theory levels (HF, MP2 and B3LYP), three basis sets (STO-3G, 6-31G* and 6-31+G*) and three population analyses (Mulliken, MK and NPA). This evaluation was, in most of cases, done for three sets of descriptors – charges from non-dissociated molecules, charges from dissociated molecules, and a combination of both these types of charges. Afterwards, we compared the accuracy of all these QSPR models and discussed the influence of all factors.

1. C. Lee, G. M. Crippen: Predicting pK_a. J. Chem. Inf. Model., 49 (2009), 2013-2033.

2. K. C. Gross, P. G. Seybold, C. M. Hadad: Comparison of different atomic charge schemes for predicting pK_a variations in substituted anilines and phenols. Int. J. Quantum Chem., 90, (2002), 445-458.

3. W. C. Kreye, P. G. Seybold: Correlations between quantum chemical indices and the pK_as of a diverse set of organic phenols. Int. J. Quantum Chem., 109 (2009), 3679-3684.

4. R. Svobodová Vařeková, S. Geidl, C.M. Ionescu, O. Skřehota, M. Kudera, D. Sehnal, T. Bouchal, R. Abagyan, H.J. Huber, J. Koča: Predicting pK_a values of substituted phenols from atomic charges, J. Chem. Inf. Model., 51 (2011), 1795-1806.