RECFIT: a microcomputer-based nonlinear regression curve-fitting package for the quantitative resolution of...

A dedicated nonlinear regression based curve-fitting packages has been developed for quantitative analysis of beta-adrenoceptor subtypes. A feature of this package is the provision to obtain initial parameter estimates using a conversational graphical technique where the user is prompted for parameter estimates, and the resulting curve is displayed over the data. Data are then fitted to a one- or two-binding site model with user-selected weighting and constraints on the parameters. A novel feature is the provision to estimate the nonspecific binding component in the assays as a parameter in the model. The printout for each model consists of the parameters and their standard deviations, estimates of the goodness of fit, an analysis of residuals, and a graph of the data points overlayed with the fitted curve. The nonlinear regression algorithm is based on that of Gauss-Newton. When unweighted, the values determined by RECFIT are essentially the same as those found using the BMDPAR programs on an ICL 2976 mainframe computer. The implementation is reasonably efficient; a typical run for a two-site fit, using nine data points and estimating nonspecific binding, took a total time of about 8 min, excluding data entry and derivation of initial estimates, 4 min for the fitting, 30 sec for graph generation, and 2-3 min for printing of the graph and data.