102
saturable high-affinity ("specific") component and one nonsaturable
("nonspecific") component, which allowed us to further check the
validity of the simple linear Scatchard method by omitting subtraction
of the 200X "nonspecific" correction. The binding model was the
following:
bt = BoF/(Kd+ F) + W- (4"6)
The actual regression equation (which is rather lengthy) was found by
substituting F = Sy By into equation (4-6) and then solving the
resulting expression for By explicitly. This regression equation (not
shown) thus does not contain F; it accepts By and Sy as pairs of input
variables and then regresses By on the relatively error-free variable
Sy, fitting the adjustable binding parameters K^, Bq, and C-j (the
asymptotic "sink" of nonspecific binding). Each input data point was
2
weighted as 1/By consistent with the assumption that the coefficient
of variation or percentage error in the measurement of By is constant
over the range of By. The resulting complete 3-parameter regression
model was then plotted in the Scatchard coordinate system for display
(e.g., figure 4-14), and each of the two components was also plotted
separately.
It would also be desirable to combine data sets from separate
isotherms in a merged, weighted nonlinear regression of By on the
relatively error-free variable Sy in a manner similar to that described
above for the simple linear "merged Scatchard" analysis. Each value of
By would be normalized by the appropriate protein concentration to the
units fmole/mg protein; the free (F) and total ligand concentration (Sy)