This is a very complex area and the answer depends very much on what you are wanting to achieve. You mention that you work in cheminformatics. I've seen several presentations from commercial companies pushing their LIMS systems in that field where the processes that go on in the lab are modelled very well in the electronic system. If that is the case, then one of those systems will be your ideal (albeit probably expensive) solution. And it won't be open source.
If you are thinking more in terms of flexible, free-form note-taking then the Marketing spin that the commercial companies put on their ELN solution falls apart. You will have to find something more open-source and adapt it to your needs.
And then you have to consider what stuff you want to share and how widely you need to share it. Science funders are often very big on sharing, but equally big on putting an onus on you to protect and exploit any IP that may arise. Obviously these are fairly conflicting requirements. I note that you have tagged your question with "collaboration" and "open-notebook science" so I guess that you are not concerned about IP (or haven't been told that you should be).
My personal preference is to use a helpdesk/ticketing system to track individual pieces of work (virtually no two tasks that I work on are the same and none fit into the conventional sample-assay-result LIMS paradigm) and then to make public data that arises from the experimental work via a separate controlled mechanism. Our "collaboration" usage thus only operates inside of our tent, but that works for us.
For the record, we use RT from bestpractical.com (http://bestpractical.com/rt/). I have no affiliation with them, but their software is useful in our hands.
