Descending lines indicate sequence. Square brackets indicate alternates. Asterisk indicates compared with 4-H compound.
Figure 1-10. Topliss decision tree. M, more active; E, equiactive; L, less active.
equations can be rapidly calculated.8 This procedure gives useful correlations by both methods in most situations and has been shown to give comparable results.
No doubt other refinements and approaches will be made to quantify structure-activity relationships of drugs to help design new agents and minimize the effort and expense of synthesizing and testing predictably inferior analogs of a lead compound. The usefulness of optimizing lead compounds will increase as experience with and understanding of quantitative methods increases, yet it should be understood that QSAR methods do not produce the lead compounds at present; ultimately, lead-generating techniques may evolve. An additional point should be made. The degree of reliability of the methods depend on the accuracy of the biological tests that, of course, have large experimental errors and variability. This may result in correlation errors. The accuracy of biological data, however, may increase in the future as new methods not using animals are developed as they already have been for some toxicological screenings.
An essentially nonmathematical approach to utilizing the basic Hansch concepts to help design drugs was developed by Topliss (1977). It has been called a decision tree (Fig. 1-10), by which a lead compound can be efficiently optimized without the use of computers. By preparing several well-chosen analogs of a lead compound, the next several
8 For an excellent brief exposition of regression analysis, see Martin, 1978, pp. 167-187.
analogs can then be planned based on the biological results of the first batch. An optimum analog can then be frequently arrived at by preparing a total of only a dozen or so compounds. One limitation of this method is that the lead structure have an unfused benzene ring.9 A scheme for nonaromatic systems also exists. The use of ti, o, and Es parameters are, of course, considered in the decisions at each step.
As one follows Figure 1-10, the first derivative of the lead compound would be its 4-C1 derivative. Since both n and o values for C are positive (Table 1-6), this is likely to result in the first analog being more active, M, than the lead. The next analog, the 3,4-dichloro, should be even more active. Further improvement should then be achieved by the 4-C1, 3-CF3 compound, and even by the 3-CF3, 4-N02 derivative. If the 4-C1 derivative were less active than the lead compound (e.g., due to steric effects), then positions other than para may be useful. Substituents with negative n and o values may be effective at the 4 position (e.g., the -OCH3). The decision tree also suggests a scheme for equiactive 4-C1 derivatives to follow; an extension of the approach to side-chain alkyl groups as may occur in esters, ketones, amines, and so on. Applications to antiinflammatory agents, diuretics, and amphetamines illustrate the utility of this method.
In summary, in designing new drugs from a discovered lead compound there must be strategy that will allow us to eliminate the synthesis and testing of structures that are likely to be ineffective or otherwise inferior from consideration. It is obvious that one cannot synthesize every conceivable derivative. Consider the hypothetical lead compound XX, where substituents can be put in the five positions indicated. Using only four substituents that are likely to offer good discrimination between steric (Es), electronic (a), and partition (k) effects, a total of 1,024 (45) compounds would have to be prepared to evaluate the "best." If we decided on expanding the experiment to seven substituents, the number of compounds needed to be synthesized would be an astronomical 16,807. It is clear that a manageable number must be arrived at. QSAR analysis and intuition must now come into play. One approach might be to apply the Free-Wilson method as an initial approach to the synthetic efforts. This will give information on the role of substituents in biological activity of interest and what functional groups to use. The Hansch analysis would then be invoked to separate the three parameters mentioned (k, c, and Es). Last, molecular-orbital (MO) calculations might then be used to help determine the electron perturbations likely to produce the desired biological activity. The MO method can be viewed as fine tuning. In a way it constitutes another way of studying the influence of substituents on a compound. MO methods have been described by Kier (1971) and others.
An alternative game plan could be to utilize the Topliss scheme as the initial approach to extract valuable information by the synthesis of a small number of compounds.
Additional approaches to help make more sense, and therefore increased predictability, out of SARs include pattern recognition. This method has been applied to the analysis of
Was this article helpful?