Molecular sorting and catalysis directed by shape selectivity have been extensively applied in porous extended frameworks for a low carbon,predictable,renewable component of modern industry.[1-4] A comprehensive understanding of the underlying recognition mechanism toward different shapes is unfortunately still missing,owing to the lack of structural and dynamic information under operating conditions.We here demonstrate that such difficulties can be overcome by state-of-the-art molecular dynamics simulations which provide atomistic details that are not accessible experimentally,as exemplified by our interpretation for the experimentally observed aggregation-induced shape selectivity for Suzuki C-C coupling reaction catalyzed by Pd particles in mesoporous silica.[5] It is found that both aggregation ability and aggregating pattern of the reactants play the decisive role in controlling the shape selectivity,which are in turn determined by the balance between the hydrophobicity and hydrophilicity of the reactants,or in other words,by the balance between the non-covalent hydrogen bonding interaction and Van der Waals forces.A general rule that allows to predict the shape selectivity of a reactant has been proposed and verified against experiments.We show that molecular modeling is a powerful tool for rational design of new mesoporous systems and for the control of catalytic reactions that are important for the petrochemical industry.