Enhancing Carbon Acid pKa Prediction by Augmentation of Sparse Experimental Datasets with Accurate AIBL (QM) Derived Values.

The prediction of the aqueous pKa of carbon acids by Quantitative Structure Property Relationship or cheminformatics-based methods is a rather arduous problem. Primarily, there are insufficient high-quality experimental data points measured in homogeneous conditions to allow for a good global model to be generated. In our computationally efficient pKa prediction method, we generate an atom-type feature vector, called a distance spectrum, from the assigned ionisation atom, and learn coefficients for those atom-types that show the impact each atom-type has on the pKa of the ionisable centre. In the current work, we augment our dataset with pKa values from a series of high performing local models derived from the Ab Initio Bond Lengths method (AIBL). We find that, in distilling the knowledge available from multiple models into one general model, the prediction error for an external test set is reduced compared to that using literature experimental data alone.

Aqueous pKa prediction for tautomerizable compounds using equilibrium bond lengths.

The accurate prediction of aqueous pKa values for tautomerizable compounds is a formidable task, even for the most established in silico tools. Empirical approaches often fall short due to a lack of pre-existing knowledge of dominant tautomeric forms. In a rigorous first-principles approach, calculations for low-energy tautomers must be performed in protonated and deprotonated forms, often both in gas and solvent phases, thus representing a significant computational task. Here we report an alternative approach, predicting pKa values for herbicide/therapeutic derivatives of 1,3-cyclohexanedione and 1,3-cyclopentanedione to within just 0.24 units. A model, using a single ab initio bond length from one protonation state, is as accurate as other more complex regression approaches using more input features, and outperforms the program Marvin. Our approach can be used for other tautomerizable species, to predict trends across congeneric series and to correct experimental pKa values.

Experiment stands corrected: accurate prediction of the aqueous pK a values of sulfonamide drugs using equilibrium bond lengths.

We show here for the first time that strongly correlated linear relationships exist between equilibrium bond lengths of the sulfonamide group and aqueous pK a values. Models are constructed for three variants of the SO2NHR group: primary benzene sulfonamide derivatives (e.g. diuretic drugs furosemide and hydrochlorothiazide), N-phenyl substituted 4-amino-N-phenylbenzenesulfonamide analogues (e.g. the sulfa antibiotic sulfadiazine) and phenylsulfonylureas (e.g. insulin secretagogue, glimepiride). In the context of these compounds, we present solutions to some of the more complex challenges in pK a prediction: (i) prediction for multiprotic compounds, (ii) predicting macroscopic values for compounds that tautomerize, and (iii) quantum chemical pK a prediction for compounds with more than 50 atoms. Using bond lengths as a powerful descriptor of ionization feasibility, we also identify that literature values for drug compounds celecoxib, glimepiride and glipizide are inaccurate. Our newly measured experimental values match our initial predictions to within 0.26 pK a units, whereas previous values were found to deviate by up to 1.68 pK a units. For glimepiride, our corrected value denotes a percentage of ionization at intracellular pH, which is only now in excellent agreement with its known therapeutic efficacy. We propose that linear relationships between bond lengths and pK a should emerge for any set of congeners, thus providing a powerful method of pK a prediction in instances where pK a data exist for close congeners, thereby obviating the need for thermodynamic cycles.

Prediction of Aqueous pK a Values for Guanidine-Containing Compounds Using Ab Initio Gas-Phase Equilibrium Bond Lengths.

In this work, we demonstrate the existence of linear relationships between gas-phase equilibrium bond lengths of the guanidine skeleton of 2-(arylamino)imidazolines and their aqueous pK a value. For a training set of 22 compounds, in the most stable conformation of their lowest energy tautomeric form, three bonds were found to exhibit r 2 and q 2 values >0.95 and root-mean-squared-error of estimation values ≤0.25 when regressed individually against pK a. The equations describing these one-bond-length linear relationships, in addition to a multiple linear regression model using all three bond lengths, were then used to predict the experimental pK a values of an external test set of further 27 derivatives. The optimal protocol we derive here shows an overall mean absolute error (MAE) of 0.20 and standard deviation of errors of 0.18 for the test set. Predictions for a second test set of diphenyl-based bis(2-iminoimidazolidines) yielded an MAE of 0.27 and a standard deviation of 0.10. The predictive power of the optimal model is further demonstrated by its ability to correct erroneously reported experimental values. Finally, a previously established guanidine model is recalibrated at a new level of theory, and predictions are made for novel phenylguanidine derivatives, showing an MAE of just 0.29. The protocols established and tested here pass both of Roy's modern and stringent MAE-based criteria for a "good" quantitative structure-activity relationship/quantitative structure-property relationship model predictivity. Notably, the ab initio bond length high correlation subset protocol developed in this work demonstrates lower MAE values than the Marvin program by ChemAxon for all test sets.

The computational prediction of Raman and ROA spectra of charged histidine tautomers in aqueous solution.

Histidine is a key component of a number of enzymatic mechanisms, and undertakes a myriad of functionalities in biochemical systems. Its computational modelling can be problematic, as its capacity to take on a number of distinct formal charge states, and tautomers thereof, is difficult to capture by conventional techniques. We demonstrate a means for recovering the experimental Raman optical activity (ROA) spectra of histidine to a high degree of accuracy. The resultant concordance between experiment and theory is of particular importance in characterising physically insightful quantities, such as band assignments. We introduce a novel conformer selection scheme that unambiguously parses snapshots from a molecular dynamics trajectory into a smaller conformational ensemble, suitable for reproducing experimental spectra. We show that the "dissimilarity" of the conformers within the resultant ensemble is maximised and representative of the physically relevant regions of molecular conformational space. In addition, we present a conformer optimisation strategy that significantly reduces the computational costs associated with alternative optimisation strategies. This conformer optimisation strategy yields spectra of equivalent quality to those of the aforementioned alternative optimisation strategies. Finally, we demonstrate that microsolvated models of small molecules yield spectra that are comparable in quality to those obtained from ab initio calculations involving a large number of solvent molecules.

The AIBLHiCoS Method: Predicting Aqueous pKa Values from Gas-Phase Equilibrium Bond Lengths.

The proposed AIBLHiCoS method predicts a given compound's pKa in aqueous solution from a single ab initio bond length only, after geometry optimization in the gas phase. Here we provide simple and predictive equations for naphthols and chemically similar biomolecules. Each linear equation corresponds to a High-Correlation Subset (HiCoS) that expresses the novel type of linear free energy relationship discovered here. The naphthol family exhibits a clear and strong relationship with the phenol family, with the "active" C-O bond always producing the highest correlations. The proposed method can isolate erroneous experiments and operate in non-aqueous solution and at different temperatures. Moreover, the existence of "active fragments" is demonstrated in a variety of sizable biomolecules for which the pKa is successfully predicted.