All of the chemical modeling work we do relies on something called the structure-property similarity principal. The fundamental idea set forth by this principal is that similar structures usually have similar properties. So, if we have some means to compare or rank the chemicals to determine how similar they are, we can use mathematical and statistical techniques to predict their properties. Many early qualitative techniques (known as structure-activity relationships or SARs) ranked chemicals based on their perceived structural similarities. Quantitative structure-activity relationships (QSARs) calculate how similar or different two structures are and weight their predictions based on this quantitative assessment.
We have developed the concept of hierarchical QSAR (HiQSAR) where topological (topostructural and topochemical), geometrical, and quantum chemical (semiempirical and ab initio) indices are used in a stepwise fashion, the more time consuming parameters are used only when the simpler parameters do not give acceptable models. HiQSAR has been applied to the prediction of property/ activity/ toxicity of molecules. In our HiQSAR analysis, we have used theoretically calculated descriptors because in the practical situation experimental properties like logP (octanol/ water), Hammett's sigma, or Taft's steric parameters, are not available for the majority of candidate chemicals both in the toxicity assessment of chemicals and drug design.
An Integrated QSAR model can be obtained by combining bioinformatics, which includes DNA descriptors, gene expression and proteomics, with cheminformatics. Overall, neither chemodescriptors alone, nor biodescriptors alone, do as well as the combination of both types of descriptors in modeling chemical toxicity.