Molecules structure, QSAR modeling validation

Once a QSAR model is constructed, it must be validated using the external test set. The data points in the test set should not appear in the training set. There are two approaches to improve the prediction accuracy for a given QSAR model. The first approach utilized the concept of "the domain of applicability," which is used to estimate the uncertainty in prediction of a particular molecule based on how similar it is to the compound used to build the model. To make a more accurate prediction for a given molecule in the test set, the structurally similar compounds in the training set are used to construct model and that model is used to make the prediction. In some cases, the domain similarity is measured using molecular descriptor similarity, rather than the structural similarity. The... 

Figure 7 illustrates the indeterminant problem in 3D-QSAR. The first case, of flexible molecules, applies in particular to the common (core) moiety if the core moiety can adopt multiple conformations, it is likely that if valid QSARs can be established for one conformer, then others may lead to equally relevant QSARs. For example, if our data set consists of tetrapeptides, then the backbone could adopt all-trans, a-helical, 3i0-helical, and (3-turn conformations. The 3D-QSAR models would be equally valid because the biological activity differences would be explained by differences in chemical structure, i.e., by side-chain variations. The relationship between the... 

Sugai er al. (1990) used a quantitative SAR (QSAR) to analyze the correlation.s between chemical structure and eye irritation in rabbits. They claimed 86.3% accuracy in classifying substances with respect to eye irritation. Barratt (1997) described an eye irritation QSAR model for neutral organic compounds. Based on the perturbation of ion transport across the cell membrane being related to dipole moments of the causative substance, the model parameters chosen were log (octanol-water partition coefficient) and the inertial axes / , and (representing the cross-sectional area of the molecule). The results were stated to provide support for the validity of the QSAR model. However, it is hoped that product safety evaluation will not be based solely on QSAR (Ballantync, 1999a),... 

---