Porous organic cages are a novel class of molecules with many promising applications, including in separation, sensing, catalysis and gas storage. Despite great promise, discovery of these materials is hampered by a lack of computational tools for exploring their chemical space, and significant expense associated with prediction of their properties. This results in significant synthetic effort being directed toward molecules which do not have targeted properties. This thesis presents multiple computational tools which can aid the discovery and design of these materials by increasing the number of synthetic candidates which are likely to exhibit desired, targeted properties. Firstly, a broadly applicable methodology for the construction of
computational models of materials is presented. This facilitates the automated modelling and screening of materials that would otherwise have to be carried out in a more labour-intensive way. Secondly, an evolutionary algorithm is implemented and applied to the design of porous organic cages. The algorithm is capable of producing cages closely matching user-defined design criteria, and its implementation is designed to allow future applications in other fields of material design. Finally, machine learning is used to accurately predict properties of porous organic cages, orders of magnitude faster than has been possible with traditional, simulation-based approaches.