Typical QSAR Model Development

The developmental process for the construction of a structure-activity model. 

In studies conducted to date, the number of compounds in this intersection has been small however, the power of including a toxicity endpoint increases the predictive power of the model when compared to models with chemical endpoints alone. 

Because a molecule is the unit of toxicity, not mass in a mg/kg, it is generally necessary and desirable to transform the LD50 and LC50 values into molar form as follows  

A variety of parameters are included into the QSAR equation. Log P is a commonly used parameter and is obtained from Medchem or estimated using the CLOGP3 computer program. Molecular weight is calculated. In interspecies models the LD50 or LC50 value is incorporated as a typical parameter. Molecular connectivity indexes, electronic charge distributions, and kappa environmental descriptors have been proven as powerful predictors of toxicity. The efficacy of these values lies in the fact that each of these parameters describes a molecule in a fashion similar to that actually seen by the molecular receptors that initiate a toxic response. Substructural keys are identified with the help of the MOLSTAC substructural key system. MOLSTAC consists of five classes of descriptors  

Identification of the longest continuous chain of atoms (excluding hydrogen) in the molecule 

---

Typically, the final part of QSAR model development is the model validation [17, 18], when the predictive power of the model is tested on an independent set of compounds. In essence, predictive power is one of the most important characteristics of QSAR models. It can be defined as the ability of a model to predict accurately the target property (e.g., biological activity) of compounds that were not used for model development. The typical problem of QSAR modeling is that at the time of the model development a researcher only has, essentially, training set molecules, so predictive ability can only be characterized by statistical characteristics of the training set model and not by true external validation. 

More typically the process of building up the QSAR models requires more complex chemical information. For a set of compounds, with known property value, the descriptors are calculated. The process of model building proceeds through a reduction of the molecular descriptors, in order to indentify the most important ones. Then, using these selected chemical descriptors and a suitable algorithm, the model is developed. Finally, the model so obtained has to be validated. 

A QSAR model is typically developed starting with a number of compounds, which are used as training set. The QSAR model is a chemometric method that extracts the knowledge relative to the population of chemical compounds with certain associated property values, such as toxicity. 

One of the most important characteristics of QSAR models is their predictive power. The latter can be defined as the ability of a model to predict accurately the target property (e.g., biological activity) of compounds that were not used for model development. The typical problem of QSAR modeling is that at the time of... 

In a typical QSAR publication it is now more usual to see some attempt to compare modelling methods against one or more datasets.Some workers have developed automated systems to do this, describing it as combi or combinatorial QSAR. ... 

To assess the types of data that may be used for QSAR analysis of skin permeabihty, refer to the reviews of Geinoz et al. (2004), Moss et al. (2002), and Vecchia and Bunge (2004b). There has been a wide variety of types of skin permeabihty data used for QSAR modeling. However, most data sets tend to be relatively small for QSAR development, typically between 5 and 20 compounds. In addition, it must be pointed out that most usable data were pubhshed from the 1960s to the 1980s. Because of the small number of individual data in data sets, it is appropriate to... 

It is appropriate to illustrate these concepts with a worked out example of a typical application of molecular quantum similarity ideas. The example addressed here involves the set of globulin bindings steroids used by Cramer et al. ° and subsequently in other studies to develop QSAR models. " This dataset has also been used by chemists in molecular quantum similarity studies and to develop quantum QSAR models. ° ° ... 

The problem addressed by the GFA is the development of accurate QSAR models that are assumed to be linear combinations of a set of features chosen from a given larger set. The given set of features could be basic molecular descriptors, principal components or complex non-linear functions of the molecular descriptors. Typically, molecular descriptors could be hydrophobicity measures, partition coefficients etc. The higher-level features which are functions of the molecular features are called basis functions. The effectiveness of the models is measured by their ability to capture parsimoniously the dependence of the desired activity on an extracted set of relevant basis functions. 

---