I'd argue that DNA is not the important thing here (and this work has nothing to do with biology): it is just a very convenient model for what chemistry can do with large varieties of interactions, as it is easy and inexpensive to synthesize and can be comparatively easily designed to have large sets of orthogonal interactions. Tile systems like this one could be built with other molecules, but it would be much harder to design.
The question in tile assembly systems is how computation can be done by the crystallization process: you could compare them, not to protein function, but to protein folding. The DNA is not being used just as instructions to a machine: there is nothing in this paper but DNA, and the computational work is done entirely through strands of DNA binding to combinations of complementary sequences on the growing crystals: the "instructions" to the computer here are actually the computer itself.