Well I guess that depends on what you're doing within you're research.
In bioinformatics knowing about, say, functional programming and parallel algorithms could be really useful where data coming out of sequencers is growing faster than Moore's law can keep up, if you want to scale across many cores as efficiently as possible.
Or graph theory, or... any number of useful techniques waiting for an application.
