3
Fig. 4-23. Affinity spectral analysis of the binding of [ H]corticosterone to the complete
ensemble of corticosterone binding sites is perfused mouse brain cytosol
(derived from the total binding data, BT, of the experiment shown in fig
ure 4-8).
Experimental conditions are described in the legend to figure 4-8, and the use of the
approximate method of finite differences is described in Methods. Distribution A is a
transformation of binding data "predicted" by the complete 3-parameter regression displayed
as the curve in figure 4-15 (i.e., binding data that have been "smoothed" by force-fit to a
specific binding model); thus, spectrum A is the representation of a single homogeneous
class of high-affinity binding sites. Distribution B is a transforamtion of total binding
data that have been smoothed by the relatively "assumption-free" cubic spline algorithm
described in Methods. In both cases the transformation method was applied to binding
functions expressed in the log By-log F coordinate system. To generate spectrum A the
curved regression line of figure 4-15 was transformed to the log By-log F coordinate system
before use of the finite differences method; spectrum B was produced by smoothing the By
data in this coordinate system with the cubic spline algorithm before transforamtion to the
affinity distribution. Both spectra have been normalized by the total binding site
concentration (B^) estimated by the regression displayed in figure 4-15. Spectrum A
(equivalent to the regression of figure 4-15): equilibrium dissociation constant = 8.8
nM; spectrum B (data smoothed by "model-free" cubic spline): equilibrium dissociation
constant K, = 5.3 nM. Method of finite differences: input spacing log K = 0.1, log
Q 1
a = 0.2. N(K) is dimensionless; dimensionality of K is M (liters/mole). Inset: Total
binding (By) data of figure 4-8 smoothed by cubic spline algorithm in the log By-log F
coordinate system (By data points not shown). The subroutine uses a statistical criterion
to smooth the data, but there is no force-fit to a specific model as in the regression
shown in figure 4-15. The "spline-smoothed" binding curve was transformed into spectrum B
by the method of finite differences (Thakur et al., 1980).