Ontologies are fundamental for organizing and representing knowledge within a specific domain. In the biomedical domain, this type of representation is essential for structuring, coding, and retrieving data efficiently. However, existing biomedical ontologies often lack coverage of all relevant concepts and relationships in the same resource. In this paper, we describe our model for semantically merging and integrating diseases, symptoms, drugs, and adverse events into a single, unified resource. We propose leveraging BioSTransformers, our developed and trained siamese neural network models on biomedical data, to discover new relevant relationships between different biomedical concepts and create novel semantic relationships between these concepts. Our objective is to build a consistent, merged ontology that serves as a valuable resource for healthcare professionals across various health-related applications. To assess and validate the newly generated relationships, we plan to use external knowledge bases such as the UMLS Metathesaurus and the Semantic Network, alongside a large language model. Initial results are encouraging and pave the way for further research.