I'm not sure the correct solution is to solely focus on mapping a database to your web stack, but rather also considering application-specific requirements.
Are you analyzing twitter feeds or other high-volume data for patterns, but don't need transactional support? Then pick something really fast.
Do you just want to store some really basic info in a few tables and its currently not an "enterprise focused" app? Then pick something cool to learn.
Maybe you are going to store data thats really important to a client, robust, needs to be transactional, and have live replication to remote hosting facilities, etc. Then maybe look at something like postgresql. It wouldn't mirror as well, but the node.js driver works pretty well and if you aren't deathly afraid of sql it gets what you want in/out pretty easily.
As for my own opinion, I think going with a newer stack like node.js (vs traditional frameworks in php/java) is adding enough "new" complexity that one shouldn't add extra layers all at once. This is a good article that discusses that:
http://nodeguide.com/convincing_the_boss.html