Just ran across a very interesting article on Ned Seeman's DNA-building machine.
This machine can be fed strands of DNA to make it reconfigure and build one of four product strands (by linking shorter strands together). I've mentioned this before. What's new in this article is that Seeman has some pretty bold plans for it:
Unlike a ribosome, Seeman's device cannot translocate. As a result, it can only make products that are pretty much the same size as itself. The team believes that future versions of the device could be able to incorporate translocation.
That would be pretty big news. It would let the device build strands of indefinite length.
"The main application I foresee is the construction of new materials that will be coded by DNA signals and then prepared by the nanomechanical system," said Seeman. "Ultimately, we will be able to make polymers and new materials of remarkable diversity and with features never seen before."
That's also big news. If he plans to build materials, and not just linear strings, then the device is (rather, its future variants are) more versatile than I thought.
A month or two ago, I sent in an X prize proposal for a machine that could build a copy of itself, as well as programmable shapes. It sounds like Seeman's plans are closer to that goal than I thought anyone would be yet. It's worth noting that this machine is made of DNA, and builds strands of DNA. It doesn't build nucleotide-by-nucleotide, but I'm not sure that's necessary in order to duplicate itself. I'm not sure if it can be done with a library of only four sequences, but with some clever design it might be possible to make pairs of sequences be the building block, for a total of up to sixteen. That would probably be enough.