My own research centers around the problem of estimating how many results a certain query will have. This is important internally in databases, as it influences what should be evaluated first, and it is important in a federated case, where you will get a huge impact if you need to do too many network requests.
I'm working on conditional selectivity, that is, if you know that you have data about a person, it is much more likely that a person has a name than a DNA checksum (yes, that kind of stuff is actually in the spec...).
However, where my work stops is where you could exploit the ontology to make further constraints, like if you know that a person has only one name, but may have several nicks, that would come in handy as well. That kind of information may exist in an OWL ontology, and if you'd like to work with me in that direction, this thesis is for you!