One interesting area both in toxicology and drug discovery is the selection of similar chemicals (commonly referred to as analogs). This can be done either intuitively or through the use of computational methods. Our group has been involved in the development of quantitative molecular similarity analysis (QMSA) methods based on computed molecular descriptors. Some research has been carried out on similarity methods based on experimental properties vis-à-vis purely theoretical descriptors.
Chemical analogs selected by various QMSA methods can be used to estimate various properties, including toxicities and bioactivities, of chemicals. QMSA methods have also been used to estimate the mode of action (MOA) of toxicants from their chemical analogs.
In many practical situations, the size of the problem is too big to be tackled by exhaustive experimental methods in the lab. Structure-property correlations are based on the theory that similar structures usually have similar properties. In a chemical structure space, chemical clusters can be identified. Clustering based on theoretical structure spaces can be used to decrease the size of the problem both for drug design and hazard assessment of chemicals.