82
the accurate derivation of KrfC from the competition displacement data
under consideration here. The ease of implementation of equation (3-36)
suggests that it (or at least equation 3-37) should always be used
instead of the Cheng-Prusoff approximation.
Although the graphical linearization of competition data is
achieved most conveniently in the recommended coordinate systems
discussed above, it can be displayed in simple modifications of any of
the popular binding plots. For example, a modified "Lineweaver-Burke"
plot (Lineweaver and Burke, 1934) that is linear with slope (and
y-intercept 1.0) can be constructed by plotting (Bq-B^J/B^ on the
ordinate with 1/F^ on the abscissa. A modified "direct linear" plot
(Eisenthal and Cornish-Bowden, 1974) may even be used to estimate by
finding the median of the abscissae of the intersections where lines
plotted for each of the individual observations in the usual "direct
linear" parameter space (F^, B^) intersect the horizontal lines having
ordinates Bq-B^.
The problem of determining the best-fitting line for the
recommended Bc/Fc vs. [Bq-Bl-Bc, (free binding sites)] plot is similar
to the problem of regression for the original one-ligand Scatchard plot
and has, in this context, been adequately discussed (e.g. Cressie and
Keightley, 1979; Rodbard, 1973; Rodbard and Feldman, 1975). In
addition, by the very nature of the definition of logit [B^/(Bl)q], the
logit-log plots of competition displacement data are quite sensitive to
error in the measurements of B^ performed at the low concentrations of
the competing ligand. Although the assumptions underlying the use of
the method of least squares (e.g., uniformity of variance,
noncorrelation of error in the independent and dependent variables) are