I am largely in the same situation as you, i.e. I used Python mostly for scientific applications, using scipy. I wanted to create a website to publish my research (see my signature for the result, but keep in
mind it's still work in progress!).
I chose CherryPy as my web framework, largely because it's simple and gets out of the way. I have found that the documentation can be somewhat lacking in certain respects but overall it's very easy to do what you want.
CherryPy does not specify a template library so I chose mako which is fast and very simple to use. The only problem I had with it was getting meaningful tracebacks when an exception was raised by the in-template code, but I was able to rectify it once I read the relevant documentation section carefully.
Modulok suggested using ORM software. ORM should not really be needed if you are aiming at scientific content for your application, you should be fine with straight SQL (many consider ORM a hindrance rather than help for any project [1], [2]). But if you do find the need for ORM then SQLAlchemy is very good and is considered pretty much a de facto standard in the Python world.
[1] https://en.wikipedia.org/wiki/Object-relational_mapping#Controversy
[2] http://goo.gl/ECNSp