With greater basis set execute superior, while the EEM QSPR models don’t show such marked variations. Similarly, the high quality of QM QSPR models depends a lot on population analysis, but EEM QSPR models aren’t influenced a lot. Namely, QM QSPR models which use atomic charges calculated from MPA, NPA and Hirshfeld PA performed really effectively, though MK provides only weak models. Within the case of EEM QSPR models, MPA performs also the ideal, but all other PAs (like MK) supply accurate benefits at the same time. The supply of the EEM parameters also didn’t have an effect on the high-quality in the EEM QSPR models substantially. The robustness of EEM QSPR models was successfully confirmed by crossvalidation. Specifically, the accuracy of pKa prediction for the test, training and full set had been comparable. The applicability of EEM QSPR models for other chemical classes was tested inside a case study focused on carboxylic acids. This case study showed that EEM QSPR models are indeed applicable for pKa prediction also for carboxylic acids. Namely, 5 from 12 of those models were able to predict pKa with R2 0.9, even though the most effective EEM QSPR model reached R2 = 0.925. Consequently, EEM QSPR models constitute a very promising method for the prediction of pKa . Their major benefits are that they are accurate, and can predict pKa values extremely promptly, because the atomic charge descriptors utilised in the QSPR model is usually obtained substantially more rapidly by EEM than by QM. On top of that, the top quality of EEM QSPR models is much less dependent on the kind of atomic charges utilized (theory level, basis set, population evaluation) than inside the case of QM QSPR models. Accordingly, EEM QSPR models constitute a pKa prediction approach which is really appropriate for virtual screening.Added file 4: Table S2. The parameters of all the QSPR models for phenols. Extra file five: Table S6. The table containing charge descriptors for all charge calculation approaches and predicted pKa values for all QSPR models (for phenols). Extra file six: Table S3. The facts about outlier molecules for phenols. Extra file 7: Table S4. The table of crossvalidation final results for phenols. More file 8: Table S5. The quality and statistical criteria of QSPR models for carboxylic acids.Abbreviations 3d: three descriptors; 4d: 4 descriptors; 5d: five descriptors; 7d: 7 descriptors; AIM: Atoms in Molecules; ANN: Artificial Neural Networks; B3LYP: Becke, threeparameter, LeeYangParr; DENR: Dynamic Electronegativity Relaxation; EEM: Electronegativity Equalization Technique; GDAC: GeometryDependent Atomic Charge; HF: HartreeFock; KCM: Kirchhoff Charge Model; LFER: Linear No cost Energy Relationships; MK: MerzSinghKollman; MLR: Many Linear Regression; MP2: M lerPlesset Perturbation Theory; MPA: Mulliken Population Analysis; NPA: All-natural Population Evaluation; PA: Population Analysis; PEOE: Partial Equalization of Orbital Electronegativity; QEq: Charge Equilibration; QM: Quantum Mechanical; QSPR: Quantitative StructureProperty Relationship; RMSE: Root Mean Square Error; SQE: Split Charge Equilibration; TSEF: Topologically Symmetric Power Function; WO: Without Outliers.Price of 133186-53-5 Competing interests The authors declare that they’ve no competing interests.Formula of 5-Bromo-4-methylthiazole Author’s contributions The concept with the study originated from JK and was reviewed and extended by RA, whilst the design and style was put with each other by RSV and SG and reviewed by JK and RA.PMID:24914310 SG and CMI collected and ready the input information. SG, OS, DS and TB performed the acquisition and postprocessing of data. Th.
