Abstract
It is shown that log P values for water–alkane or water–cyclohexane partitions, and the corresponding Δ log P values when used as descriptors for blood–brain distribution, as log BB, yield equations with very poor correlation coefficients but very good standard deviations, S from 0.25 to 0.33 log units. Using quite large data sets, we have verified that similar S‐values apply to predictions of log BB. A suggested model, based on log P for water–dodecane and water–hexadecane partition coefficients, has 109 data points and a fitted S = 0.254 log units. It is essential to include in the model an indicator variable for volatile compounds, and an indicator variable for drugs that contain the carboxylic group. A similar equation based on water–chloroform partition coefficients has 83 data points and a fitted S = 0.287 log units. We can find no causal connection between these log P values and log BB in terms of correlation or in terms of chemical similarity, but conclude that the log P descriptor will yield excellent predictions of log BB provided that predictions are within the chemical space of the compounds used to set up the model. We also show that model based on log P(octanol) and an Abraham descriptor provides a simple and easy method of predicting log BB with an error of no more than 0.31 log units. We have used the Akaike information criterion to investigate the most economic models for log BB. © 2009 Wiley‐Liss, Inc. and the American Pharmacists Association J Pharm Sci 99: 2492–2501, 2010
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have