Thanks everyone for replying so quickly!
I've had a look at the SDK, and tried out the examples, but I haven't written any code yet (as much as I would like to). Because it would take longer than we have to write anything meaningful, the focus of our project is more on evaluating the potential for GPGPU in the Life Sciences with applications that are available now.
Specifically, we're looking at usability, levels of effort and reward for different approaches/platforms, applicability to our clusters, etc. There are GPU accelerated programs with applicable to Life Sciences written in CUDA (NAMD, HOOMD, etc.), but I haven't been able to find anything similar/equivalent in ATI Stream.
I've had a little luck with ACML-GPU, but I thought I should ask here if anyone knows of anything else. OpenMM sounds good too, but the code for that isn't available yet.