rnEstimation of brain uptake of solutes and drugs would be useful in designing novel therapeutic agents. Given the problems associated with fragment-based models, including the need for strict additivity, we used AM1-derived molecular parameters in attempting to generate an equation for predicting the log of the brain-to-blood concentration ratios (log B:B) for a collection of compounds. Sequential fitting of the data using 10 AM1-derived variables indicated that dipole generated the most predictive one-term equation, dipole and log P , the best two-parameter system, dipole, log P , and molecular weight produced the optimal three-term model, and dipole, log P , molecular weight, and HOMO energy resulted in the best four-parameter paradigm. These systems were statistically similar to those generated using general solvation equations, but with the use of fewer parameters. An expanded parameter set may generated models of even better predictability. 0 1996 John Wiley & Sons, Inc.