| Method |
The binding constants and structural components of 200 drugs and enzyme inhibitors have been used to calculate the average binding energies of 10 common functional groups. As expected, charged groups bind more strongly than polar groups, which in turn bind more tightly than nonpolar groups. In regression analyses for which the intercept has been
arbitrarily fixed (in this case, to a known value), standard
statistical parameters for evaluating the fit (e.g., the
multiple correlation coefficient) and for determining the
accuracy of derived regression coefficients (e.g., t tests) are
not meaningful. An alternative measure of the accuracy of researchers' intrinsic binding energies (coefficients in eq 4 in article) was
therefore obtained by running regression analyses on
random subsets of the initial 200-compound data base to
monitor variations in the derived coefficients. The ranges
of Ex (Intrinsic binding energy) values for six separate sets of 100 compounds are
shown in Table. In terms of percentage errors, the
smaller coefficients are less well defined than the larger
values, but, in general, the ranges in Ex are fairly small.
The values in Table 2 are certainly more tightly defined
than have been previous estimates of intrinsic binding
energies. |