Accounting concepts

The physiochemical properties of drugs play an important role in the absorption and bioavailability of the drug substance and thus, to its effectiveness to reach to its target. Hence, these properties must be predicted at the preclinical stage of the drug development to avoid the chances of its failure at the later stage. The development in the field of computational biology has significantly reduced drug failures. In-silico approach to drug discovery and development now makes it possible to apply various techniques at the early stage of drug discovery to virtually screen the molecules which don’t appear to have good absorption/solubility properties. A lot of solubility models have been devised so far based on the experimentation as well as molecular descriptors. This paper elaborates the descriptive background of solubility prediction as well as various models developed in the recent years.

1. 
   Introduction:
   

In 1991, about 40% of drug candidate attrition was caused by adverse pharmacokinetics (PK) and bioavailability[2]. The development in the field of computational biology significantly reduced drug failures. In last five years, more than 158 drugs were approved by FDA for various diseases. QSAR and ADME-Tox like properties have been evaluated before clinical testing. However, poor Solubility and bioavailability is still a major factor that leads to drug attrition. The drug aqueous solubility and membrane permeability are the major factors affecting drug`s bioavailability[3].

2. 
   Prediction of Solubility of substance:
   

The Biopharmaceutics Classification System is a system to differentiate the drugs on the basis of their solubility and permeability. It is a guide for predicting the intestinal drug absorption provided by the U.S. Food and Drug Administration[6]..

According to the Biopharmaceutics Classification System, drug substances are classified as follows:

There are mainly four types of solubility as described in the relevant literature:

The aqueous solubility models can be categorized into two types, those correlated with experimentally determined properties and those not[7]. There are a number of ways in which the models can be classified. Even they can be classified on the basis of the characteristics of descriptors. They can even be classified in terms of various methods implemented to derive the model. The most specific classification can be:

1. 
   Solubility models based on experimental measurements.
   
2. 
   Solubility models devised using molecular descriptors.
   

For a given solid state and solvent, the solubility S is almost exclusively dependent on the intermolecular adhesive interactions between solute-solute, solute-solvent and solvent-solvent molecules. That is to say the crystal packing, cavitation and solvation energy determine the intrinsic solubility of a compound[7]. A number of models have been devised so far, based on the equations generated on the various experimental properties including melting point, octanol-water partition coefficient, molecular weight, free energy of solvation, vapor pressure of pure substances, boiling point etc.

• 
  Molecular Descriptors:
  

By applying the mathematical and/or logical operations on such structures, we can devise the short term that serve as the self explanatory describer of the elements and structure of the molecules. Such derivations are known as “Molecular Descriptors”.

Molecular descriptors are formulated by applying various concepts and principles from different theories/areas like Information theory, Quantum Chemistry, Organic Chemistry, Molecular Graph Theory etc.

• 
  Molecular descriptors and QSAR Modeling:
  

Several years back, the solubility models were being developed using huge datasets. There was a tendency among developers to use larger data set to develop the model. Gradually, some new developments have started the trend of designing and customizing new datasets according to the need and requirements.

Despite the vital importance of the solubility data of different compounds from chemical, biochemical, pharmaceutical and industrial data sets, there is still a lack of fundamental solubility and mass-transfer data available in the literature to facilitate the development of solubility models[11].