Two-stage median-based methods for characterizing enzyme inhibition.

Abstract

Non-parametric linear regression is used for inhibition diagnostics and parameter estimation based on the velocity equation v = sKsV/(1 + sKS + iKI + isKIS) = sV(i)/(K(i) + s) = W(s)/(L(s) + i). From velocities measured as a function of inhibitor level i at different constant substrate levels s, median-based estimates of 1/W(s) and L(s)/W(s) are derived. Different diagnostic secondary plots are introduced, which take special shapes at limiting inhibition types. Two-stage median-based methods are presented for fitting the four-parameter equation to the observed velocities. These methods cause only little computational effort in contrast to one-stage median-based estimation which requires medians to be taken of solutions of impracticably large numbers of linear equation systems. Performance of the new methods is compared with that of the least-squares fit by application to simulated velocity data. For velocity data containing outliers, the median-based estimates are superior to the least-squares ones, whereas for normally distributed data the reverse is true.