Comparing Literature Solubility Models

The reviews published by Bergstrom [12] and Delaney [13] in 2005 and Dearden [15] in 2006 give good overviews of the predictive solubility literature. 

But the main problem with the multitude of solubility models is how to compare them The only way to obtain an objective judgment of their predictive power is to test each model using a common external test set. For this purpose, Dearden made a compilation of results from 21 models, whose authors published results of a common test set comprising 20 drugs and pesticides and one PCB, as it was initially used by Yalkowsky and Banerjee to evaluate their own model. A summary of the most homogeneous results from these models is reproduced in Table 4.1. 

This comparison is very useful, but one must keep in mind that 20 molecules in a test set does not represent a large chemical space, and although it constitutes a comparative test, it does not prove that the models are predictive for any drug-like molecule, as we shall discuss later. Dearden widens his comparative study to 17 commercially available programs able to predict solubility. The comparison was done 

Author Year of publication Compounds in training set Compounds in test set Std of training set Std of test set 

Du-Cuny et al. [21] have used a large data set of247 3 compounds likely to be more drug-like as they are from their own employer s pharmaceutical company collection. The originalities of their work are 

---

Water solubility (thermodynamic) prediction, based on various literature datasets [32, 33], is comparable to other models [29]. This model was confirmed with GPSVS. 

The Flory-Huggins and Wilson equations for the activity coefficients of the components of the mixed solvent were employed to correlate 32 experimental data sets regarding the solubility of drugs in aqueous mixed solvents. The results were compared with the models available in literature. It was found that the suggested equation can be used for an accurate and reliable correlation of the solubilities of drugs in aqueous mixed binary solvents. It provided slightly better results than the best literature models but has also the advantage of a theoretical basis. 

The literature and commercial companies abound with computational solubility models. Many data sets have been studied, with many different descriptor sets, and using a multitude of statistical methods. It appears that diverse drug-like data sets are often predicted by our best methods with an RMSE of 0.8-1 log unit. This compares with an error in replicate measurements of approximately 0.5 log unit. A common view is that there is still room for improvement in the computational modeling of solubility. There are a number of suggestions that the quality control of the ideal data set is still lacking. This may be true for some literature data set compilations, but it is... 

The extraction of toluene and 1,2 dichlorobenzene from shallow packed beds of porous particles was studied both experimentally and theoretically at various operating conditions. Mathematical extraction models, based on the shrinking core concept, were developed for three different particle geometries. These models contain three adjustable parameters an effective diffusivity, a volumetric fluid-to-particle mass transfer coefficient, and an equilibrium solubility or partition coefficient. K as well as Kq were first determined from initial extraction rates. Then, by fitting experimental extraction data, values of the effective diffusivity were obtained. Model predictions compare well with experimental data and the respective value of the tortuosity factor around 2.5 is in excellent agreement with related literature data. 

The paper is organized as follows first, an equation for the activity coefficient of a solute at infinite dilution in a binary nonideal mixed solvent (Ruckenstein and Shulgin, 2003) is employed to derive an expression for its solubility in terms of the properties of the mixed solvent. Second, various expressions for the activity coefficients of the cosolvents are inserted into the above equation. Finally, the obtained equations are used to correlate the drug solubilities in binary aqueous mixed solvents and the results are compared with experimental data and other models available in the literature. 

HT-solubility/permeability First, solubility is determined at four pH values by comparing the concentration of a saturated compound solution with its dilute, known as the concentration. The filtered, saturated solution from the solubility assay is then used as input material for the membrane permeability determination. The permeability assay is a parallel artificial membrane technique whereby a membrane is created on a solid support, PAMPA. The two artificial membranes presented here model the GIT and the BBB. Data are presented for control compounds, which are well documented in the literature and exemplify a range of solubility and membrane permeability. The advantages of the combination method are (/) reduction of sample usage and preparation time, ( /) elimination of interference from compound precipitation in membrane permeability determination, Hi) maximization of input concentration to permeability assay for improved reproducibility, and (/v) optimization of sample tracking by streamlining data entry and calculations. BBB permeability ranking of compounds correlates well with literature CNS activity. 

The PROPERTIES option provides measured values taken from the literature or values calculated from QSAR models for 10 molecular descriptors that relate to the chemicals physical and chemical properties molecularweight, parachor, molar refraction, molar volume, boiling point, vapour pressure, water solubility, log P, heat of vapourization, and log S (see Lyman et al. 1982, for a detailed discussion of these parameters). The user may alter the value for any parameter in order to see how that may affect values of other parameters in the system. The option also generates connectivity indices which are highly useful in describing and comparing molecular structural characteristics (Kier and Hall 1976). 

Solubility of Gaseous Mixtures Comparison with Experimental Data. The experimental determination of the solubility of gas mixture components in a solid polymer is a very complicated task relative to the measurement of solubility of a pure gas or vapour. Indeed, very few mixed gas sorption data are available in the technical literature. To compare with predictions of the NELF model, only the experimental data for the solubility of GO2/C2H4 mixtures in PMMA at 35°C (4, 75, 76) will be considered in this article. Since the volume dilation produced by the gas mixture is not available, the analysis of the data is confined to the low pressure range, where the swelling of the glassy polymer may be neglected. Indeed, in the calculation which follows the non-equilibrium solubility is estimated from the NELF model assuming... 

---