Prediction of distribution of neutral, acidic and basic structurally diverse compounds between blood and brain by the nonlinear methodology.

The methodology for predicting the distribution of compounds between Blood and Brain, i.e. their brain/blood partition coefficients (logBB values), was studied using a nonlinear regression analysis in this work. The equations were established on the basis of the different states (neutral, cationic and anionic) of the compounds distributing into the three dominating composition (lipid, protein and water) of the brain. The equations bear strong fitting and predictive power for the distribution of compounds (total set: n=160, r=0.906, s=0.326; training set: n=139, r=0.908, s=0.320; testing set: n=21, r=0.903, s=0.297), and can describe the distribution of the different states of the compounds in three compositions of brain. The compounds in the dataset contained many different types, such as drug molecules, small structure-simple molecules, carboxylic acids and also alkaloids. Therefore the equations were very useful and instructional for the prediction of the compound distribution into the brain and blood. Finally, the percentages of the amount of a compound in lipid, protein and water in brain were calculated using the model, such subdivision will be very useful in drug research and discovery. By an analysis of the percentages a conclusion can be obtained that a well distributed drug is mainly affected by distribution of lipid and protein.