QSPR models of a certain type and percentages of QSPR models whose R2 values are in a specific intervalQSPR model Average R2 Interval of R2 R2 0.9 0.9 R2 QSPR model Typical R2 Interval of R2 R2 0.9 0.9 R2 0.85 0.85 R2 0.8 0.85 0.85 R2 0.eight 3d EEM 0.876 11 83 6 3d EEM WO 0.911 83 17 0 EEM based models 0.900 63 35 2 5d EEM 0.913 94 six 0 3d QM 0.929 78 six 17 5d QM 0.951 83 17 0QM based models 0.940 81 13 6Table 4 Average R2 amongst experimental and predicted pKa for all QSPR models working with atomic charges calculated by a particular combination of theory level and basis set, or by a certain population analysisQSPR model Theory level and basis set Population evaluation HF/STO3G B3LYP/61G MPA NPA Hirshfeld MK CHELPG AIM 3d EEM 0.878 0.889 0.889 0.884 0.842 0.867 0.870 0.3d EEM WO 0.919 0.917 0.917 0.907 0.884 0.914 0.886 0.5d EEM 0.918 0.918 0.918 0.907 0.905 0.914 0.906 0.3d QM 0.952 0.967 0.967 0.959 0.904 0.845 0.853 0.5d QM 0.966 0.972 0.972 0.968 0.948 0.896 0.909 0.Only QSPR models employing MPA have been included within this evaluation. Only QSPR models using B3LYP/61G were included in this evaluation.SvobodovVaekovet al. Journal of Cheminformatics 2013, five:18 a r a http://www.jcheminf.com/content/5/Page 9 ofHF/STO3G/MPA/3dB3LYP/631G/MPA/3dB3LYP/631G/NPA/3dB3LYP/631G/MK/3dcalculated pKacalculated pKacalculated pKacalculated pKa0 2 4 6 eight 10experimental pKaexperimental pKaexperimental pKaexperimental pKaHF/STO3G/MPA/5dB3LYP/631G/MPA/5dB3LYP/631G/NPA/5dB3LYP/631G/MK/5dcalculated pKacalculated pKacalculated pKacalculated pKa0 two 4 6 eight 10experimental pKaexperimental pKaexperimental pKaexperimental pKaSvob2007_chal2/3dChaves2006/3dBult2002_npa/3dBult2002_mk/3dcalculated pKacalculated pKacalculated pKacalculated pKa0 2 4 six 8 10experimental pKaexperimental pKaexperimental pKaexperimental pKaSvob2007_chal2/3d WOChaves2006/3d WOBult2002_npa/3d WOBult2002_mk/3d WOcalculated pKacalculated pKacalculated pKacalculated pKa0 two four 6 eight 10experimental pKaexperimental pKaexperimental pKaexperimental pKaSvob2007_chal2/5dChaves2006/5dBult2002_npa/5dBult2002_mk/5dcalculated pKacalculated pKacalculated pKacalculated pKa0 two four 6 eight 10experimental pKaexperimental pKaexperimental pKaexperimental pKaFigure 2 Correlation graphs.Methyl 4-chloro-3-methylpicolinate Purity Graphs showing the correlation in between experimental and calculated pKa for chosen QSPR models.Table 4). This really is in agreement with our earlier findings [24], and it can be explained by the truth that 61G is usually a a lot more robust basis set than STO3G. Nonetheless, the difference is not marked in the case of EEM QSPR models.Influence of population analysisEleven EEM parameter sets had been published for B3LYP/631G with six various population analyses (see Table 1).330645-87-9 custom synthesis Thus it is actually straightforward to analyze the influence with the population analysis around the predictive power on the resulting QSPR models (see Table 4).PMID:23937941 We discovered that MPAand NPA give the ideal QM models, although MK and CHELPG (PAs according to fitting the atomic charges towards the molecular electrostatic possible) supply weak QM models. Our outcomes are in agreement with previous studies [22,24]. QM QSPR models determined by AIM predict pKa with accuracy comparable to MPA and NPA. Inside the case of EEM QSPR models, we did indeed come across that MPA provided the most beneficial models, but a lot of the other population analyses gave comparable benefits. This confirms previous observations that the EEM method isn’t capable to faithfully mimic MK charges [63]. On the other hand,SvobodovVaekovet al. Journal of Cheminformatics 20.
