Do not use clustalW for the philogenetic tree. The tree produced by clustalW in NOT a phylogenetic tree, it just clusters together similar sequences based on matches-mismatches. Use Phylip or PAUP or similar programs instead. The program must be designed to produce phylogenetic tree.

For the root ancestor you need an outgroup. In the case of a gene family it might be difficult to decide what is the right outgroup. Must be a gene that you know diverged from all the others first. I have used Phylip and it allows you to set the outgroup as "root" (the first ramification)

If you are clustering genes in human it might make sense to add a bacteria homologue (if it exists) or a member of a "sister" family. Or another kinase, possibly a related one. A mouse homologue (orthologue) most likely diverged from the human one when the ancestor already had a few of them.

Also, remember that, unless you have "fossils" you can't find the root ancestor. All human hexokinase are "alive" and you cannot say that all of them came from one of them. They all came from one that does not exist anymore. The dendrogram will show you the phylogeny, i.e. the history of how they duplicated. On nodes there will be "unknown ancestors" that do not exist anymore.

Hope this is helpful

Stefano covered the most important bits, and you really shouldn't use the clustal trees to make any inferences, but you should also be aware than almost any tree viewing program will allow you to root a tree you've already estimated.

I usually end up using treeview or the R libary ape, but there are a tonne of programs that will work