Abstract：The challenging problem of modeling blood-brain barrier partitioning is approached through molecular structure. Here a QSAR model was developed for in vivo logarithm value of blood-brain concentration ratio (LogBB), by the method of Bayesian Regularization Neural Network (BRNN). The model consists of eight structure descriptors which elucidate topological properties,partition coefficient and size of molecules. The model for a set of 52 chemicals is validation through a use of external prediction set (10 in 52 chemicals, R2=0.974, MSE=0.0172), the quality of validation statistics support the claim that this model is much better than multi-variance linear regression and may be use for estimation of LogBB value for drug and drug-like molecules.
