We describe a new ontology-driven semantic data integration approach for post-genome biology studies. Here, a view-based global schema can be automatically generated by merging RDF schemas from local databases. The semantic inconsistency of the merged schema is resolved by the creation of 'RDF ontology maps'. Data querying capability is accomplished with a virtual data repository, in which a D2RQ-based 'relational-to-RDF' map is developed to link schema to the relational database backend. With sample RDQL queries, we demonstrate that our approach significantly simplifies the retrieval of human protein interaction data from different databases containing hundreds of thousands of records.