Yes, this! I work in bioinformatics, and while relational DBMSes are used by a large number of projects, the problem you face with experimental data is that its organisation is non-obvious and relationships between different bits of data are not apparent. This means that RDBMSes aren't a natural fit. One of my colleagues has been experimenting with graph oriented databases (specifically neo4j (http://neo4j.org/)), an approach which has interesting intersections with declarative programming. In the near future I think much of use will come from being a scientific "data geek", utilising the kind of skills described by Deepak Singh (blog: http://mndoci.com/).