pharmine

A QSAR/qSAR model is a mathematical model which approximates the relationship between the pharmacological activity of a compound and its structure-derived physicochemical and structural features. The two main objectives of QSAR/qSAR modeling are to allow prediction of the pharmacological activity of a not yet biologically tested, but chemically characterized compound and to extract clues as to which molecular characteristics of a compound are important for the pharmacological activity. The term “QSAR model” is used to refer to quantitative models (regression problems), and the term “qSAR model” will be used to refer to qualitative models (classification problems). There are variants of QSAR like QSPR (for chemical properties), QSPkR (for pharmacokinetic properties), QSTR (for toxicological properties).

The initial step is the collection of relevant pharmacological activity data and the elimination of low quality data that are likely to affect the quality of the model. The next step is the selection of representative compounds into a training set and a validation set to calibrate and evaluate the QSAR/qSAR model respectively. Molecular descriptors are then computed for representing the physicochemical and structural properties of the compounds studied and those that are redundant or contained little information are removed prior to the modeling process. A machine learning method or statistical method is then used to develop a model that relates the pharmacological activity to the physicochemical and structural properties of the compounds. During a modeling process, optimization of the essential parameters of the machine learning methods or statistical methods and the selection of relevant descriptor subsets are conducted simultaneously. The optimum set of parameters and descriptor subset are used to construct a final QSAR/qSAR model, which is subsequently subjected to various validation methods to ensure that it is valid and useful.

This entry was posted by Yap Chun Wei on Thursday, February 21st, 2008 at 11:17 pm and is filed under Data mining, Pharmacy, Tutorial. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.