Good question! I thought of the kinase classfication based on the similarity of lists of the known small molecule inhibitors. Since kinase inhibitors are not the most speciffic actives in the world, such classification may help to search off-targets among kinases faster, for example. But, it would be not so easy, to make a usefull classification this way; given the sparsity of the data on kinase inhibition, not to mention their complexity.