Other QSAR Approaches

45 — 65%. In evaluating these figures, it must be considered that even in unbalanced groups, i.e. independent of the number of active and inactive analogs being included in the test set, there is a 50% probability of success ( ) of an unbiased blind guess. 

It is claimed that CASE differs from other techniques in being completely automatic and by learning directly from the crude data, selecting its own descriptors from the practically infinite number of possible structural assemblies and creating an ad hoc dictionary without human interference [524]. While this statement is against all prior experience with automated approaches, even from a critical point of view it cannot be ruled out that the CASE approach may for the first time be an approximation of artificial intelligence to the medicinal chemist s intuition and skill. 

In the topological pharmacophore methods [527 — 531], e.g. LOGON [527, 528, 530], LOGANA [528 — 530], and EVAL [531], Free Wilson-type indicator variables 

Magee [544, 545] combined the hyperstructure concept with the strategies of Hansch analysis and the mixed approach (chapter 4.3). As only several atoms or groups of a molecule modulate biological activity, each position of the hypermolecule 

It seems too early to judge on the real suitability of neural nets for QSAR studies further investigations which compare classical structure-activity analyses and results from neural networks e.g. [570]) are required to evaluate the scope and limitations of neural nets. Some problems of neural networks, e.g. the design of the network, lack of convergence, chance correlations, and overtraining of the network, have been discussed and critically commented [562, 567 — 570]. 

---

Finally, unlike most other QSAR approaches, the cr-moment approach allows for the graphical visualization of the logarithmic partition coefficients as local surface properties of the solutes,... 

With the development of accurate computational methods for generating 3D conformations of chemical structures, QSAR approaches that employ 3D descriptors have been developed to address the problems of 2D QSAR techniques, that is, their inability to distinguish stereoisomers. Examples of 3D QSAR include molecular shape analysis (MSA) [26], distance geometry,and Voronoi techniques [27]. The MSA method utilizes shape descriptors and MLR analysis, whereas the other two approaches apply atomic refractivity as structural descriptor and the solution of mathematical inequalities to obtain the quantitative relationships. These methods have been applied to study structure-activity relationships of many data sets by Hopfinger and Crippen, respectively. Perhaps the most popular example of the 3D QSAR is the com-... 

The literature of the past three decades has witnessed a tremendous explosion in the use of computed descriptors in QSAR. But it is noteworthy that this has exacerbated another problem rank deficiency. This occurs when the number of independent variables is larger than the number of observations. Stepwise regression and other similar approaches, which are popularly used when there is a rank deficiency, often result in overly optimistic and statistically incorrect predictive models. Such models would fail in predicting the properties of future, untested cases similar to those used to develop the model. It is essential that subset selection, if performed, be done within the model validation step as opposed to outside of the model validation step, thus providing an honest measure of the predictive ability of the model, i.e., the true q2 [39,40,68,69]. Unfortunately, many published QSAR studies involve subset selection followed by model validation, thus yielding a naive q2, which inflates the predictive ability of the model. The following steps outline the proper sequence of events for descriptor thinning and LOO cross-validation, e.g.,... 

The results of the studies will be summarized. Details of the QSAR analyses are or will be published elsewhere, including intercorrelation matrices of the steric parameters mentioned. But relevant conclusions from e.g. intercorrelations will be dicussed. At this moment the STERIMOL method has been applied successfully in about 50 publications often with better results than other steric approaches, including MTD and MTD, especially in series with few substituent positions. A recent example is our study of DDT analogs. Brown et al. (9J analysed a series of 21 derivatives using the van de Waals (Vw) volumes as steric parameters. In Table I the equations are given in which the steric parameters are compared. 

Any QSAR method can be generally defined as an application of mathematical and statistical methods to the problem of finding empirical relationships (QSAR models) of the form ,- = k(D, D2,..., D ), where ,- are biological activities (or other properties of interest) of molecules, D, P>2,- ,Dn are calculated (or, sometimes, experimentally measured) structural properties (molecular descriptors) of compounds, and k is some empirically established mathematical transformation that should be applied to descriptors to calculate the property values for all molecules (Fig. 6.1). The goal of QSAR modeling is to establish a trend in the descriptor values, which parallels the trend in biological activity. In essence, all QSAR approaches imply, directly or indi-... 

Hansch analysis and other classical QSAR approaches evaluate the QSAR model based on the correlation of rows of compounds with known activities (dependent variables) to columns of parameters (independent variables). For this reason, classical QSAR is sometimes called 2D QSAR. CoMFA is an example of 3D QSAR because lead analogues are modeled and analyzed in a virtual three-dimensional space. The value of both methods ultimately hinges on how well experimental and calculated activities correlate (Figure 12.2) and how well the model predicts the activity of compounds not included in the training set. 

Probabilistic methods. Other QSAR-like probabilistic approaches have also been developed for compound database mining. Binary QSAR (BQ) is discussed here as an example (Labute 1999). BQ is based on Bayes theorem of conditional probabilities ... 

While QSAR approaches (in all their forms) are by far the most common ways to capture and explore SAR trends, a number of other approaches are possible. Although they are not quantitative, they can be useful as idea generators. ... 

Based on the earlier work of Meyer and Overton, who showed that the narcotic effect of anesthetics was related to their oil/water partition coefficients, Hansch and his co-workers have demonstrated unequivocally the importance of hydrophobic parameters such as log P (where P is, usually, the octanol/water partition coefficient) in QSAR analysis.28 The so-called classical QSAR approach, pioneered by Hansch, involves stepwise multiple regression analysis (MRA) in the generation of activity correlations with structural descriptors, such as physicochemical parameters (log P, molar refractivity, etc.) or substituent constants such as ir, a, and Es (where these represent hydrophobic, electronic, and steric effects, respectively). The Hansch approach has been very successful in accurately predicting effects in many biological systems, some of which have been subsequently rationalized by inspection of the three-dimensional structures of receptor proteins.28 The use of log P (and its associated substituent parameter, tr) is very important in toxicity,29-32 as well as in other forms of bioactivity, because of the role of hydrophobicity in molecular transport across cell membranes and other biological barriers. 

A QSAR approach based on a set of methods that combines molecular shape similarity and commonality measures with other - molecular descriptors both to search for similarities among molecules and to build QSAR models [Hopfinger, 1980 Burke and Hopfinger, 1993], The term molecular shape similarity refers to molecular similarity on the basis of a comparison of three-dimensional molecular shapes represented by some property of the atoms composing the molecule, such as the van der Waals spheres. TTie molecular shape commonality is the measure of molecular similarity when conformational energy and molecular shape are simultaneously considered [Hopfinger and Burke, 1990]. 

The physicochemical properties most commonly studied by the QSAR approach have been described above, but other properties have also been studied. These include dipole moments, hydrogen bonding, conformation, and interatomic distances. However, difficulties in quantifying these properties limit the use of these parameters. 

---