Given the demand for applying knowledge graphs and ontologies in diverse kinds of applications and contexts, it is highly desirable to assure their structural flexibility, such that the same semantic content could be expressed via slightly different patterns and idioms. The key to such flexibility is the capability to execute structural transformation at the level of ontologies, which can be then propagated to instance knowledge graph, too. While ontology transformation based on symbolic patterns has been around as task for over a decade, its wider adoption was hindered by the overwhelming demand for manual tuning of its individual steps. Nowadays, with the increasing power of pre-trained Large Language Models (LLMs), this hindrance could however be lifted. The research presented in the planned PhD thesis will aim to exploit LLMs as a complement to symbolic patterns within the transformation process. As the key tasks for LLMsapplication in this process we foresee: 1) transformation pattern applicability assessment; 2) generation of new entity name and definition; 3) generation of training examples for the previous task, as a kind of bootstrapping. In the paper, preliminary results are also being reported for the second task.