How is it different from previous computers?

1) What is your team's main contribution?
We show the first ever nanoscale programmable autonomous computing machine, and in particular the first such machine in which the input, output, software, and hardware are all made of biomolecules.

2) How does your work fit into the body of work on DNA computing, and what is different about it?
It is a continuation of designs for DNA-based Turing machines (references 4-8 in the Letter to Nature) in the sense that it is the first working implementation of a specialized Turing machine (that's what a finite automaton is). It is different from most experimental work on DNA computers (refs 9-15) in that we do not attempt to compete with silicon computers by solving difficult problems faster. Our long term goal is to construct nanocomputers that solve biological problems by interacting with their biochemical environment in a useful way, which does not necessarily mean solving complex computational problems but rather producing the right molecular output at the right place and time based on the molecular input and the "software" program. We solve simple problems, autonomously, and in a biochemical environment.

3) Could explain the historical background of your work?
One can trace two lines of research in DNA computing. One, started by Adleman in 1994 (reference 9 in the Letter to Nature), is to use the power of DNA manipulation to compete with electronic computers in solving complex combinatorial problems. The other is to build nanoscale molecular computers from biomolecules. This line of work started by the conceptual design of a molecular Turing machine by Bennett in 1982 (reference 4), with traces of these ideas already in Bennett's seminal paper on reversible computing. Our work belongs to Bennett's line of work. While there were several designs for such computers (e.g. references 4-8), I believe we report on the first ever nanoscale programmable autonomous computing machine, and of course the first such machine in which the input, output, software and hardware are all made of biomolecules. I should note that we did not implement a full-fledged Turing machine, but rather a programmable 2-state 2-symbol finite automaton.

Some steps in this direction are acknowledged in the Letter. For example Seeman (ref 16) shows that specially designed DNA, if cooled down slowly, can form preprogrammed patterns. This has yet to be applied to computation. Hagia (ref 17) shows how DNA can change its internal state in a preprogrammed sequence, but this requires controlled cycling of temperature and therefore it is not fully autonomous, and the sequence of state changes does not depend on any external input, so it is not quite an automaton in the mathematical sense of the word.

Our relationship to other previous experimental work (Adleman's line, references 9-15) is best explained by saying that we pursue different goals. The stated goal of most previous experimental work in DNA computing (refs 9-15) was to use the small scale of DNA molecules and the inherent parallelism of DNA manipulation in the lab to compete with silicon computers in solving difficult computational problems faster. Our long term goal is to construct nanocomputers that solve biological problems by interacting with their biochemical environment in a useful, program controlled, way, which does not necessarily mean solving complex computational problems but rather producing the right molecular output at the right place and time based on the molecular input and the "software" program.


Back