Brains are not computers, and do not process information like computers.
The trees from CS you are talking about are ways to run linear information stores through a central processing unit. Even with machines that can sort of do these operations in parallel, they are still very much series computations.
Brains, on the other hand, are massively parallel, and information is encoded in a very high dimensional space, not linearly. There are no "bits" in a sequence to sort through. Information does not just move from "storage" to some "processing" part of the brain, it is constantly being processed at all stages in all places. When a memory is active, that information is linked to all the other related information physically because neurons are connected to each other. That is, when you think of a "cat", you might retrieve a memory of a specific cat you saw in a specific place because there is a physical connection between some of the same neurons that are active when you think of "cat" and the neurons encoding that engram of the specific cat memory. No "search" is necessary; every extremely high-dimensional brain state is followed by another extremely high-dimensional brain state. When you "search" your memory for cats, you are just biasing these future states to be those which involve neurons that were involved in the past during experiences of cats.
I think the most attractive models for understanding how brains index information are attractor networks. There are in silico models and implementations of these sorts of networks, but without the parallel hardware of a biological brain they aren't particularly efficient.
