The computational design of new and interesting inorganic materials is still an ongoing challenge. The motivation of these efforts is to aid the often difficult task of crystal structure determination, to rationalize different but related structure types, or to help limit the domain of structures that are possible in a given system. Over the past decade, simulation methods have continuously evolved towards the prediction of new structures using minimal input information in terms of symmetry, cell parameters, or chemical composition. So far, this task of identifying candidate structures through an analysis of the energy landscape of chemical systems has been particularly successful for predominantly ionic systems with relatively small numbers of atoms or ions in the simulation cell. After an introductory section, the second section of this work presents the historical developments of such simulation methods in this area. The following sections of the work are dedicated to the introduction of the building unit concept in simulation methods: we present simulation approaches to structure prediction employing both primary (aggregate of atoms) and secondary (aggregate of coordination polyhedra) building units. While structure prediction with primary units is a straightforward extension of established approaches, the AASBU method (automated asssembly of secondary building units) focusses on the topology of network-based structures. This method explores the possible ways to assemble predefined inorganic building units in three-dimensional space, opening the way to the manipulation of very large building units (up to 84 atoms in this work). As illustrative examples we present the prediction of candidate structures for Li(4)CO(4), the identification of topological relationships within a family of metalphosphates, ULM-n and MIL-n, and finally the generation of new topologies by using predefined large building units such as a sodalite or a double-four-ring cage, for the prediction of new and interesting zeolite-type structures.