Within the AAA superfamily, most classifications into families and subfamilies are obvious.
An odd subfamily are the ARC proteins found only in Actinomycetales (e.g., A2126A from Mycobacterium) consisting of the only eubacterial member which does not belong to the family of zink proteases. They probably have been aquired by a horizontal gene transfer, which also transmitted the genes of the 20 S proteasomal core subunits located in the vicinity of the ARC genes.
The mei-1/Meiosis family seems to be the most heterogenous and least defined, further sequence data or functional analysis may result in a subdivision in the future. Recent additions to the database indicate that the mei-1 family is not confined to eukaryotes as a few archebacteria code for proteins clustering with mei-1 and relatives.
F11A10.1 from Caenorhabditis and yeast YTA7 form a cluster, however, F11A10.1 has a single copy of the AAA box, and YTA7 a second, though very poorly conserved one.
The decision to use the amino proximal copy of the AAA box for the proteins of the Homotypic Fusion (CDC48) family for the AAA tree construction introduces another problem. The family consists of a "core" of proteins with two well conserved copies, found in all kinds of eukaryotes and in archebacteria, but seems to "bud" potential new families (I will have to restructure the tree in a few 100 million years, I suppose:-) with only one well conserved copy. The PEX family might have branched off the CDC48 family recently (in terms of evolution); as you can see below, the carboxy proximal boxes from the CDC48 and PEX families form a joint branch in the tree. YTA7, AFG2, smallminded from Drosophila, the newest AAA member from yeast on chromosome XII (the baby has no official name yet), and K04G2.3 from Caenorhabditis are candidates for future families. The first four have a well conserved amino proximal AAA box and therefore are positioned correctly in my AAA tree. Judging from its poorly conserved amino proximal AAA box, K04G2.3 looks like a far relative only, but its well conserved carboxy terminal box clusters nicely with the corresponding (carboxy proximal) CDC48 boxes (as you can see below; however, in the AAA tree using the amino proximal CDC48 boxes, it looks like an orphan). The maybe oddest finding is that the closest relative of the worm K04G2.3, MTCY22G10.32c, is found in a eubacterium, Mycobacterium tuberculosis.
BTW, in contrast to the AAA tree this is not a clickable map with links to the corresponding descriptions (access the far relatives from their list).
I am glad about corrections, additions (new sequences, new functions), suggestions, questions, requests, and praise.