Solubility predictions, model

Recently, several QSPR solubility prediction models based on a fairly large and diverse data set were generated. Huuskonen developed the models using MLRA and back-propagation neural networks (BPG) on a data set of 1297 diverse compoimds [22]. The compounds were described by 24 atom-type E-state indices and six other topological indices. For the 413 compoimds in the test set, MLRA gave = 0.88 and s = 0.71 and neural network provided... 

The accurate prediction of the aqueous solubility of drugs and drug-like compounds is much further away from a satisfactory solution because the existing QSAR- or group-contribution-based solubility prediction models exhibit quite limited predictive power for new drug classes. Why is it that the extremely important problem of the prediction of aqueous solubility much less solved than the prediction of less important partition coefficients The answer is that the development of prediction models for logarithmic partition coefficients is much simpler, because the molecule X under consideration only acts as a solute at infinite dilution in the two phases. Hence the task is only to calculate the free energy of... 

Tutorial Developing Models for Solubility Prediction with 18 Topological Descriptors... 

In order to develop a proper QSPR model for solubility prediction, the first task is to select appropriate input deseriptors that are highly correlated with solubility. Clearly, many factors influence solubility - to name but a few, the si2e of a molecule, the polarity of the molecule, and the ability of molecules to participate in hydrogen honding. For a large diverse data set, some indicators for describing the differences in the molecules are also important. 

Despite the plethora of data in the scientific literature on thermophysical quantities of substances and mixtures, many important data gaps exist. Predictive capabilities have been developed for problems such as vapor-liquid equihbrium properties, gas-phase and—less accmately—liquid-phase diffusivities, aud solubilities of uouelectrolytes. Yet there are many areas where improved predictive models would be of great value. Au accrrrate and rehable predictive model can obviate the need for costly, extensive experimental measurements of properties that are critical in chemical manufactming processes. 

Eros D, Keri G, Kovesdi I, Szantai-Kis C, Meszaros G and Orfi L. Comparison of predictive ability of water solubility QSPR models generated by MLR, PLS and ANN methods. Mini Rev Med Chem 2004 4 167-77. 

Ktihne, R., Ebert, R-U., Schtitirmann, G. Model selection based on structural similarity - method description and application to water solubility prediction. f Chem. Inf Model. 2006, 46, 636-641. 

As a key first step towards oral absorption, considerable effort has been directed towards the development of computational solubility prediction [26-30]. However, partly due to a lack of large experimental datasets measured under identical conditions, today s methods are not sufficiently robust for reliable predictions [31]. Nonetheless, further fine-tuning of these models can be expected since high-throughput data have become available for their construction. 

Huuskonen, J., Salo, M., Taskinen, J., Aqueous solubility prediction of drugs based on molecular topology and neural network modeling, J. 

If the pore-mechanism applies, the rate of permeation should be related to the probability at which pores of sufficient size and depth appear in the bilayer. The correlation is given by the semi-empirical model of Hamilton and Kaler [150], which predicts a much stronger dependence on the thickness d of the membrane than the solubility-diffusion model (proportional to exp(-d) instead of the 1 Id dependence given in equation (14)). This has been confirmed for potassium by experiments with bilayers composed of lipids with different hydrocarbon chain lengths [148], The sensitivity to the solute size, however, is in the model of Hamilton and Kaler much less pronounced than in the solubility-diffusion model. 

Taking into account wide utilization of fullerenes in many aspects of nanotechnology (Prylutskyy et al., 2003 Colbert and Smalley, 1999 Wood, 2004 Sivaraman et al., 2001 Faulon and Carla, 2003 Faulon et al., 2004 Danauskas and Jurs, 2001 Kiss et al., 2000 Franco et al., 2007) one can see the necessity of development of predictive models for their physicochemical properties. Among such properties is their solubility. Solubility of fullerene is not only an important technological characteristic, but also an ecological characteristic (Franco et al., 2007). 

According to Gasteiger et al. [59], the correlation coefficient r between bioavailability and HIA is 0.498 for 161 compounds. This conclusion inspires us to propose the use of aqueous solubility, descriptors of HIA models, and some rule-based descriptors to predict first-pass metabolism, to model bioavailability. Another research direction for the prediction of oral bioavailability is to develop separate prediction models for different components involved in oral bioavailability, including passive transcellular transport, paracellular transport, carrier-mediated transport, and first-pass metabolism, and then integrate them together. At present, the development of an integrated model is really difficult or even impossible because the predictions for some mechanisms involved in oral bioavailability are really unreliable. 

Wang, J., Hou, T., Xu, X. Aqueous solubility prediction based on weighted atom type counts and solvent accessible surface areas. J. Chem. Inf. Model. 2009, 49, 571-81. 

Bergstrom, C.A.S., Wassvik, C.M., Norinder, U., Luthman, K. and Artursson, P. (2004) Global and local computational models for aqueous solubility Prediction of drug-like molecules. Journal of Chemical Information and Computer Sciences, 44, 1477-1488. 

Zebrafish embryo assay results were compared to the ToxCast in vitro assay features from the predictive model of developmental toxicity (50). A majority of the features were significant between the zebrafish data and predictive models, despite the fact that the zebrafish assay did not correlate with global developmental toxicity defined by species-specific ToxRefDB data. The top 15 chemicals predicted to be developmental toxicants and bottom 15 chemicals predicted not to be developmental toxicants varied in their endpoint responses and logP values. Padilla et al. (35) noted that chemical-physical characteristics could limit the amount of chemical seen by the embryo due to poor solubility or poor uptake. This may be the reason that a majority of the bottom 15 chemicals with no zebrafish embryo activity had logP values less than 1.0. The bottom 15 chemicals with zebrafish embryo activity could almost exclusively be characterized by the negative predictors of the species-specific developmental toxicity models, which may be indicating that these predictors have differing roles between mammalian and zebrafish development. 

There are several properties of a chemical that are related to exposure potential or overall reactivity for which structure-based predictive models are available. The relevant properties discussed here are bioaccumulation, oral, dermal, and inhalation bioavailability and reactivity. These prediction methods are based on a combination of in vitro assays and quantitative structure-activity relationships (QSARs) [3]. QSARs are simple, usually linear, mathematical models that use chemical structure descriptors to predict first-order physicochemical properties, such as water solubility. Other, similar models can then be constructed that use the first-order physicochemical properties to predict more complex properties, including those of interest here. Chemical descriptors are properties that can be calculated directly from a chemical structure graph and can include abstract quantities, such as connectivity indices, or more intuitive properties, such as dipole moment or total surface area. QSAR models are parameterized using training data from sets of chemicals for which both structure and chemical properties are known, and are validated against other (independent) sets of chemicals. 

Malcolm et al. (14) and Thurman et al. (15) noticed that the adsorption of solutes onto XAD-8 macroreticular resin could be predicted by means of a linear correlation between the logarithm of the capacity factor and the inverse of the logarithm of the water solubility of each compound. Their investigation, however, was limited to approximately 20 selected organic compounds in individual aqueous solutions. By comparing the results shown in Table II and the water solubility properties of each model compound used in this study (see Table I), it appears that the predictive model could serve for a first estimate of the recovery of multisolute solutions at trace levels. However, low recoveries and the erratic behavior of several compounds included in this study suggest that additional factors need to be considered. 

Bergstrom, C. A., C. M. Wassvik, U. Norinder, K. Luthman, and P. Artursson. 2004. Global and local computational models for aqueous solubility prediction of drug-like molecifle hem. Inf. Comput. Sci 44 1477-1488. 

---