In object-oriented databases, relationships are generally maintained explicitly. The partial result of a retrieved object can be used to efficiently retrieve related objects. Instead of optimizing joins as in relational database systems, pointer chasing is optimized in object-oriented database systems. Further, semantics inherent in the object-oriented database, like superclass-subclass relationships and composite-component relationships between object classes, must be realised. In this paper, we describe our initial result in query optimization in an object-oriented database system. Semantic query transformation is used to preprocess the query. semantically optimized query is then translated into a query evaluation plan which comprises method invocations that can be evaluated directly by the system. In the process of query evaluation plan generation, initial results tend to show that a one source query plan is almost optimal. A prototype based on this design has been completed and some results from a simulation study on this prototype are also reported in this paper.