Test Series in QSAR

The statistical analysis of data requires a proper design of experiments to prove or disprove a certain hypothesis which has been formulated in advance. From the viewpoint of a puritanical statistician most QSAR analyses are forbidden , because they are retrospective studies and, in addition, many different hypotheses (i.e. combinations of independent variables) are tested sequentially. Indeed, many problems arise from the application of regression analysis in ill-conditioned data sets. Only in later stages of lead structure optimization are certain hypotheses, e.g. on the influence of more lipophilic, electronegative, polar, or bulky substituents in a certain position, systematically tested, now fulfilling the requirements for the proper application of statistical methods. 

Parameter focusing is a related technique, developed by Magee [634], Different 2D plots of physicochemical properties are drawn to find out which parameter combination separates active and inactive compounds to the largest extent. Therefore, the method can be applied, in contrast to QSAR analyses, also to qualitative data. Cluster significance analysis (chapter 5.3) was developed from this approach. 

A manual method was proposed by Topliss [639] as a modification of his operational schemes a larger number of substituents is selected in the first step to derive the dependence of biological activities on n (linear and nonlinear) and a with a minimum number of analogs. This latter approach has been criticized because of collinearity and unbalanced spanning of the parameter space [640]. 

The different approaches proposed by Topliss should not be understood as rigid schemes they are strategies which have to be adjusted to each problem. A recent review [403] lists more than 50 references where the Topliss methods have been applied, mostly in medicinal chemistry. It was shown that optimum activity would have rapidly been reached in many series of compounds in accordance with the Topliss scheme [641] on the other hand, there are at least some examples where the Topliss method failed [403, 633]. 

All stepwise approaches have the disadvantage that iterative synthesis and testing are extremely time-consuming strategies. In most cases a brief study of the parameter table and the correlation matrix provides a QSAR practitioner enough details on the parameter spanning and on the interrelation of the parameters (with the exception of multiple correlations). 

---

The use of statistical design strategies to select test series in QSAR studies is particularly valuable, as it permits to rationalize both the advancement of knowledge and the investment of resources. In fact, by this approach it is possible to select only a few molecules to be synthesized and tested, but in such a way that they contain the widest information, and therefore they permit to derive reliable QSAR models while saving a lot of the resources presently required for a single study. 

Pleiss MA, Unger SH. The design of test series and the significance of QSAR relationships. In Ramsden CA, editor, Comprehensive medicinal chemistry. Vol. 4 Quantitative drug design. Oxford Pergamon Press, 1990. p. 561-87. 

The series editors wish this book a wide distribution. It is up to the reader to find out which of the properties and descriptors might be most suitable for describing the data. However, the early warnings by Corwin Hansch, John Topliss, and others should not be forgotten make your model as simple as possible test and include only a few parameters try to achieve an understanding of your model use a test set to check the external predictivity of your model. Molecular descriptors are powerful tools in QSAR studies - but their abuse may lead nowhere. May this book further contribute to their selective and proper use ... 

V. Austel, in QSAR and Strategies in the Design of Bioactive Compounds, J. K. Seydel, Ed., VCH Publishers, FRG, 1984, pp. 247—250. Design of Test Series by Combined Application of 2 Factorial Schemes and Pattern Recognition Techniques. 

Finally, very recently Pallavicini et al., in continuation of a previous study on ortho-monosubstituted compounds, designed and synthesized a series of 2-[(2-phenox-yethyl) aminomethyl]-l,4-benzodioxanes ortho-disubstituted at the phenoxy moiety [99]. The disubstituted analogues were tested for their binding affinities at the three oq-AR subtypes and for the 5-HT1A-R. The affinity values of the new compounds were compared with those of the enantiomers of the 2,6-dimethoxyphenoxy analogue, the well-known oq-AR antagonist WB-4101 (Scheme 8.1), and of the ortho-monosubstituted derivatives. The results suggested some distinctive aspects in the interaction of the phenoxy moiety of monosubstituted and disubstituted compounds with the cqa-AR and the 5-HTiA receptors. A classical (Hansch) QSAR analysis was applied to the whole... 

Optimization of biological properties in a series of miticidal and mite ovicidal 2-aryl-l,3-cycloalkanediones, Ia,b, and enol esters, II, was achieved through analog synthesis and testing supported by the development of quantitative structure/ activity trends during the course of the project. QSAR equations developed during an initial phase provided the basis for both... 

In the early 1970 s potent acaricidal activity in 3-aryl-4-hy-droxy-coumarins. III, and 2-aryl-l,3-indanediones, IV, and their enol esters was discovered at Union Carbide (1,2). Extensive synthesis and screening of analogs in the dione and enol ester series led to the field test candidates Va and Vb, active against both motile forms (mite) and eggs. Va and Vb embodied certain structural and physicochemical features shown by published QSAR studies (2,3) to be important in determining the level of activity in the 1,3-indanedione series (1) the dimension D2 across the aryl ring must be near 7A and symmetry is preferred (2) the... 

Further studies revealed that in the first tests very coarse particles of these two derivatives had been used, while the testing of the other derivatives had been performed with fine particles. This focused attention for the first time on the great importance of particle size in the evaluation of the benzoyl-phenyl ureas. The series discussed included Du 19111 the first compound found, but the analyses indicated that other derivatives were more active. From two or three of the most active compounds PH 60-38 (ill) was chosen because it was found to be the one which could be synthetized most economically on an industrial scale. Consequently PH 6O-38 was taken for preliminary development both in the USA and in Europe. It is interesting to note that PH 6O-38 was one of the compounds retested after the QSAR studies. 

---