Large and flexible molecules, like proteins, carbohydrates and other natural products exhibit numerous degrees of freedom. It is generally quite difficult to find the most stable, and biochemically most active conformations and their paths of interconversion. Application of Molecular Dynamics (MD) has been established as one of the most important theoretical techniques to study the conformational space of these molecular systems. The sheer amount of highly redundant and multidimensional data produced by MD calculations constitutes a key problem of this method. This thesis will present the first application of artificial neural networks of the Kohonen type (which closely resemble the organization of the visual cortex) to investigate this wealth of data. Thus, two-dimensional representations of the conformational hypersurface are obtained, which, like topological maps, visualize conformational minima and interconversion pathways in an intuitive way. The three-dimensional molecular structure also influences important molecular properties, such as the aromaticity in cyclic, conjugated rings (annulenes). 1964 the theoretical chemist E. Heilbronner suggested that annulenes, which are twisted in themselves by 180° in a Möbius topology, should be aromatic with 4n electrons, thus, breaking the Hückel (4n+2)-rule. Despite of numerous attempts, until very recently no experimental proof of this hypothesis was found. For large annulenes, a numerous conformations and configurations exist (e.g. [30]annulene). To find the "Möbius needles" in the "annulene haystack" an algorithm was developed and implemented in a program which automatically generates isomers and analyses them for Möbius topology and aromaticity. Indeed, several stable Möbius isomers have been discovered which can be considered as interesting synthetic targets.