Selecting Substituent Sets

The simplest design just lets the D-optimal algorithm select a given number of substituents from the candidate set. This would be a purely statistical diversity design it is unbiased, but has linle else to recommend it. The next level of complexity is to select a fixed number of substituents for inclusion by hand, based on pharmacological or synthetic ideas, and then let the D-optimal 

The designs are then sorted by the diversity scores (log jX Xj), as in Table 3,to critique the tradeoff between bias, synthetic ease, and diversity. The object is to find synthetically feasible biased designs with high diversity and pharmacophore focus, and low molecular weight. Unbiased design 1 was made by taking all 35 substituents from the All candidate bin. Its score of 150.9 is used 

Design No. Score Pharm. Prot. Drug Valid. Low MW All 

---

The symbols were defined in Section IV.B.) A full discussion of the ortfio-effect as revealed in this work would be inappropriate here. We must restrict ourselves to the more limited task of indicating the role of o-N02. We discuss first the work involving alcohols as solvents. To apply the extended Hammett equation, i.e. to determine the regression coefficients a, fi and

intercept term h, it is first necessary to select a set of substituents which can be expected to be well-behaved . Particular problems for or and v may be caused by conformational effects, and internal hydrogen-bonding may occur... 

The most obvious generalization about the equations of Table la is that since all are linear in log P, more active members of each set could be made simply by increasing lipophilic character. However, in undertaking such a study it would be most desirable to select substituents so that the effects of electronic and steric factors could be evaluated simultaneously. 

The combinatorial docking tools PRO SELECT [121] and CombiDOCK [122] are based on the incremental construction method. In both approaches, a library is formed by a template (or core) molecule with a set of attachment points to which one out of a predefined set of substituents can be connected. The template is then positioned in the active site without considering the substituents. Starting from a few orientations of the template, the substituents are placed into the active site of the protein independently. In case of PRO SELECT, substituents are then selected based on score and additional criteria like 2D similarity and feasibility of synthesis. CombiDOCK calculates a final score for whole library molecules by combining fragment scores. 

D-optimal design methods [389, 403, 655, 656] calculate the determinant of the variance-covariance matrix its value is largest for substituent sets with maximum variance and minimum covariance (linear and multiple correlation) in their physicochemical properties. An information theory approach, which leads to comparable results, has been proposed by Herrmann [657]. In some other approaches synthetic accessibility has been included as an additional selection feature [656, 658, 659]. 

Nine substituents (Set A H, Me, Bu F, Cl, Br, I, CF3 and CN) were selected as a basic set for the following qualities (i) symmetry and simplicity, (ii) freedom from conformational effects, (iii) lack of a large resonance effect, (iv) lack of any marked tendency to form hydrogen-bonds with the reaction centre. It proved possible to expand the list from 9 to 18 by making reasonable assumptions about the conformations of certain substituents, thus enabling them to be placed on the Or and v scales [Set B Set A - - Ft, PF, OMe, OEt, OPh, SMe, S02Me, CH2Ph, (CH2)2Ph]. Correlations based on Set B turned out to be superior to those based on Set A. 

For the purpose of characterizing chemicals for designing and selecting training sets for QSAR analyses, a values can be retrieved from tabulations (e.g. Perrin, Dempsey and Serjeant, 1981 Hansch, Leo and Taft, 1991). If several substituents are present on a parent structure, the sum of the corresponding cr values is used. Especially for c>-substituents, specific (rvalues for the given chemical class have to be used. 

With a lead in hand, the first problem is the selection of the initial set of substituents to make. Ideally, this step involves selection of a set of substituents that represents the whole universe of factors involved in the activity of the lead compound and does so with the smallest number of compounds possible. If this step is not done well, there is a high probability that the resultant QSAR analysis will not be successful. For example, Martin and Panas present evidence that poor series design can be the cause of a failure to find a QSAR model for activity. Conversely, adoption of strategies for the proper selection of substituent sets may be all that is needed to dramatically increase the efficiency of the design process and ensure the development of QSAR models. Indeed Martinhas suggested It is the opinion of many in the QSAR field that if the only contribution of QSAR to medicinal chemistry were to emphasize the importance of objective Series design, this in itself would be a major contribution to increased efficiency. ... 

The criteria used for selection of that initial set are to a great extent dependent on the overall strategy being followed. Indeed, we can divide the strategies based on the method used to select substituents as indicated in Table 1. 

The Ideal Substituent Set Should Cover the Selected Factor Space as Completely as Possible... 

Since we wished to optimize this series with a minimal expenditure of our synthetic resources, the sequential simplex technique (SSO) was selected (11,12). This strategy is very resource efficient, requiring only n + 1 compounds to start the optimization, where n is the number of physiochemical parameters used to describe the characteristics of a substituent. We selected pi to account for lipophilicity and field (F) and resonance (R) were used to describe the electronic effects of each substituent (14). Verloop s Sterimol parameters (H), minimum van der Waals radius (B] ) and length (L) were selected to describe the size of the substituent. Using cluster analysis, we selected a set of six substituents that cover physiochemical parameter space well (15). These are-listed in Figure 6. 

In order to allow any multiple chlorination of the biphenyl skeleton, the user may define an atom list (eonsisting of hydrogen and chlorine atoms) and substitute all H-atoms by this list. One may click on the drop-down selection box behind the element icons, select the options Generics. .set the user-defined atom to A1 and quit by the OK button. As a result this atom selection is active for the subsequent drawing steps. After this atom list is drawn ten times as the ten substituents, its composition has to be defined by clicking the A, icon on the left-hand side of the structure editor and by selecting H and Cl in the periodic table (Figure 5-16). 

---