In the Web of data, an increasing number of knowledge graphs (KGs) are concurrently published, edited, and accessed by human and software agents. Their wide adoption makes key their construction, matching, and usage. KG construction can rely on knowledge extraction from various types of data (e.g., text, tables). Matching consists in identifying equivalent, more specific, or somewhat similar units within and across KGs. This task is crucial since concurrent publication and edition may result in coexisting and complementary KGs. These graphs can also be mined to discover new and useful knowledge units. Their knowledge can support applications such as explainable AI or recommender systems. However, all these tasks require to face the inherent heterogeneity of KGs, e.g., in terms of granularities, vocabularies, and completeness. Additionally, scalability issues arise due to their increasing size and their combinatorial nature. In this talk, I will present my work on knowledge graph construction (from tabular data, from other knowledge graphs), matching, and usage (explainable AI and recommender systems). Part of this work is applied to the biomedical domain of pharmacogenomics. Throughout all my presentation, I will highlight the advantages of domain knowledge in the form of ontologies associated with knowledge graphs. Indeed, when considered, such domain knowledge allows to face heterogeneity and scalability issues.