IV. LogP Comparisons, continued

Figures 1(a) and (b) show the correlation between theory and experiment graphically for the generic set of parameters and the user-trained set. It is evident that user training yields a narrower distribution that is closer to the y = x ideal agreement.

Figure 1. Comparison of logP_exp vs logP_calc calculated by the ACD/Labs method using A. generic parameters and B. user-trained parameters for 18 compounds from Ref.[1].

Fitting each of the theory vs. experiment distributions to the linear equation

yields values of the slope, intercept, correlation coefficient (R) and standard deviation (SD) as shown in Table 3. From this table, as well as from Figure 1, it is clear that the ACD/Labs algorithm works best for this set of compounds when a user-trained set of parameters is implemented.

Table 3. Comparison of the experimental vs. calculated linearities of the ACD/LogP method, before and after system training with compounds (III) and (IV) has been implemented.