Prediction of blood-brain partitioning using Monte Carlo simulations of molecules in water

The brain-blood partition coefficient (log B B) is a determining factor for the efficacy of central nervous system acting drugs. Since large-scale experimental determination of log B B is unfeasible, alternative evaluation methods based on theoretical models are desirable. Toward this direction, we propose a model that correlates log B B with physically significant descriptors for 76 structurally diverse molecules. We employ Monte Carlo simulations of the compounds in water to calculate such properties as the solvent-accessible surface area (SASA), the number of hydrogen bond donors and acceptors, the solute dipole, and the hydrophilic, hydrophobic and amphiphilic components of SASA. The physically significant descriptors are identified and a quantitative structure-prediction relationship is constructed that predicts log B B. This work demonstrates that computer simulations can be employed in a semi-empirical framework to build predictive QSPRs that shed light on the physical mechanism of biomolecular phenomena.