I see how the differences could be frustrating, especially if you’re after a ‘like for like’ data model for your topics. Yet a 6 level tree seems daunting from a retrieval perspective. (I mean, a user trying to find something.)
If each node had 3 branches, then you’ve got 729 endpoints in a 6 level tree. And if you don’t really have 700+ endpoints, then why the need for all the branching?
We don’t know your user base, of course. They might be data scientists! But it might be that having the users build the tree dynamically has left you with a majority of categories with only one post.
A flatter tree structure might actually increase engagement because users find it easier to know where to post and where to find interesting discussions.
My suggestion is to think about how you could redesign your categories to make it easier to post/find content, before even picking a platform.
(I’m making a ton of assumptions about your information and your users, of course. Apologies if I’ve made some wildly incorrect assumptions. 
)
PS - I believe the suggestions above is to have you set the categories in Discourse (up to two levels deep) and have your users create ‘tags’ for the topics they post to make it easy for them to retrieve their posts using the tag terms.) So they aren’t building a tree, but they are building a tag cloud that can be searched. Perhaps that’s the best of both worlds?