Bringing together, in a unique immersive environment, visualization and analysis of scientific and complex data requires a thorough approach in order to fulfill scientists' specific expectations. Such an approach needs to consider the highly heterogeneous nature of data, the dynamic interactions between experts and data, and the large amount of data involved in scientific studies. Whereas small and static scientific datasets can quickly be deciphered thanks to standard immersive tools such as 3D visualization software packages, bigger and dynamic datasets exceed the analytical capacity of these tools, requiring an efficient platform for their manipulation. Through the example of the structural biology field we discuss the need for an approach based on a high-level definition of the content (scientific data) and the context (immersive environments and interfaces). Our design is illustrated by a platform for dynamic and intelligent representation of data to the user. The data hierarchical classification will provide new ways to interact with the data via intelligent and direct relationships between them. This approach is based on the semantic definition of all the concepts manipulated in the virtual environment, either abstract or concrete, which allows for an adaptive and interactive experience of both visualization and analysis.