My two main requirements for the site are related to degrees of separation and graph matching (given two graphs, return some kind of similarity score).
My first thought was to use MySql to do it, which would probably work out okay for storing how I want to manage 'friends' (similar to Twitter), but I'm thinking if I want to show users results which will make use of graphing algorithms (like shortest path between two people) maybe it isn't the way to go for that.
My language of choice for the front end, would be Python using something like Pylons but I haven't committed to anything specific yet and would be willing to budge if it fitted well with a good backend solution.
I'm thinking of using MySQL for storing user profile data, neo4j for the graph information of relations between users and then have a Python application talk to both of them.
Maybe there is a simpler/more efficient way to do this kind of thing. At the moment for me it's more getting a suitable prototype done than worrying about scalability but I'm willing to invest some time learning something new if it'll save me time rewriting/porting in the future.
PS: I'm more of a programmer than a database designer, so I'd prefer having rewrite the frontend later rather than say porting over the database, which is the main reason I'm looking for advice.