In Silico Prediction of Solubility

PSA Polar surface area (dynamic or static - calculated from 2D or 3D 

RMSE Root mean square error of predictions - average deviation of the 

Solubility is widely regarded as one of the most difficult physical properties to predict. But when building predictive solubility models, or in fact any model, one needs to answer a number of questions what solubility measures are required to be modeled Do we have a suitable data set on which to build a computational model What descriptors and what modeling methods should be used How accurate are the models required to be What is the influence of the domain of applicability Do we know when good predictions have been made In this chapter, we will highlight recent research, in an attempt to answer these key questions in solubility modeling. 

Solubility is a fundamental compound quality indicator and plays a critical role in many aspects of drug research. Although most research has focused upon modeling solubility in water or aqueous buffered solutions, solubility in other milieu may be equally important. 

A number of mathematical models have been developed to describe the interplay of solubility and these physiological parameters to model dmg absorption. The most simplistic model is the maximal absorbable dose (MAD) calculation. The MAD calculation combines the amount of dmg that can dissolve to form a saturated solution in water equal in volume to the small intestinal volume, with an estimate of the absorption rate and the small intestinal transit time. The maximal absorbable dose is then related to the dose required to achieve the desired therapeutic effect [2], If the estimated MAD is much greater than the predicted dose to achieve a therapeutic effect, this can give confidence enough to take the dmg toward clinical use. Predictions of aqueous solubility may then be useful in predicting the extent of absorption in man. 

---

Lobell M and Sivarajah V. In silico prediction of aqueous solubility, human plasma protein binding and volume of distribution of compounds from calculated pKa and AlogP98 values. Mol Divers 2003 7 69-87. 

Bergstrom CAS (2005) In silico predictions of drug solubility and permeability two rate-limiting barriers to oral drug absorption. Basic Clin Pharmacol Toxicol 96 156-161. 

Engkvist, O. and Wrede, P. High throughput, in silico prediction of aqueous solubility based on one- and two-dimensional descriptors./. Chem. Inf Comput. Sci. 2002, 42, 1247-1249. 

Goller, A. FI., M. Flennemann, J. Keldenich, and T. Clark. 2006. In silico prediction of buffer solubility based on quantum-mechanical and FIQSAR- and topology-based descdpGtnam. Inf. Modefl6 ... 

Lobeh M. Advances in the in-silico prediction of aqueous solubility from structure. Cerius2 User Group Meeting. Cerep, Paris, 2001. 

Dearden JC. In silico prediction of aqueous solubility. Expert Opin Drug Discov 2006 1 31-52. 

In this respect, the in silico prediction of the thermodynamic mixing behavior of different polymer-drug/excipient mixtures is of central interest. A common approach to cope with this problem is the calculation of the solubility parameters according to Hildebrand or Hansen [9-12], which is standard in the development of polymer mixtures [13]. The use of highly developed force fields as the basis of any MD simulation software enables the calculation of solubility parameters with accuracy comparable to those measured experimentally by inverse gas chromatography [14], and an increasing number of other statistical quantitative property relationships between simulated and experimental values are established [15-18]. 

QMPRPlus was used to generate in silico estimates of log P, aqueous solubility, and human jejunal permeability from 3D molecular structures. The predictive... 

In silico methods that are able to predict quantitative aspects of the interaction of a substrate with P-gp would be of great value. So far, modeling was applied mainly to lock-key-type reactions taking place in aqueous solution. The structural diversity and lipid solubility of P-gp substrates and the fact that their encounter with the transporter takes place in the lipid membrane and not in aqueous solution are new challenges for in silico predictions. Since all in silico models are based on experimental data, we first provide a short introduction to various P-gp assays and discuss their underlying principles (18.2). Secondly, we summarize the different in silico approaches (18.3), and, lastly, we discuss the parameters that are most relevant for the different in silico models (18.4). 

---