Machine Learning Search for Suitable Structure Directing Agents for the Synthesis of Beta (BEA) Zeolite Using Molecular Topology and Monte Carlo Techniques

3D-QSAR algorithms allowed a significant advance in the identification of new drugs and established the concept relating the effects of shape on binding of steroids to carrier proteins. Zeolites do share this “common principle” by the structure directing effects of occluded organic molecules that fit conveniently in the zeolite micropores formed during the synthesis. Host–guest van der Waals interactions are of crucial importance and their calculation allows to test numerically the concept above. Molecular topology (MT) has emerged as a powerful in silico approach for the identification of new molecules with tailor-made properties. MT descriptors (hundreds of them) include topological as well as structural and chemical information of molecules. By using experimental information (either qualitative or quantitative), it is possible, by using different statistical techniques, to select a small number of descriptors whose combination gives an algorithm able to predict a particularly biological or physicochemical property. It is very important to select appropriate descriptors to assess successfully the performance of organic molecules as structure directing agents (OSDA) for the synthesis of zeolites. Monte Carlo techniques will be used to test the agreement between zeo-OSDA van der Waals stabilization energies predicted by machine learning and linear regression algorithms.