286 CATALYZING INQUIRY
A related problem is the lack of programmability of current models. Even if experimental verifica-
tion of Turing-complete biomolecular computing can be achieved, individual runs must still be care-
fully tuned to a specific instance of a specific problem, much like the hardwiring of the first generation
of electronic computers. Worse yet, the sequences of biomolecules synthesized for a particular
biomolecular computation are usually consumed or destroyed during the computation. For a replica-
tion of the experiment, even with the same dataset, much of the entire process of setup must be
repeated. If a different dataset or a different “program” is run, then other steps must be included, such
as designing the set of sequences to be used as “words” of the computation and determining the set of
enzymes and concentration levels necessary to correctly identify, mark, destroy, and read out the
appropriate strands of nucleic acids. The ability to formulate a problem of any generality in terms that
map onto a set of chemical processing lab procedures is likely an essential aspect of DNA computing,
but it is not at all clear today how such formulations can occur in general.
Finally, the most significant challenge is the high bar that DNA computation will have to surpass to
gain wide acceptance. Moore’s law is expected to continue unabated for at least a decade, resulting in
petaflop machines by 2015. Additionally, biomolecular computation is not the only radical technique in
town; quantum computation, various other applications of nanotechnology, analog computing, and
other contenders may turn out to offer more favorable performance, programmability, or convenience.
These challenges are quite significant and possibly decisive. Len Adleman himself was pessimistic
about the prospect of general computation in a 2002 paper: “Despite our successes, and those of others,
in the absence of technical breakthroughs, optimism regarding the creation of a molecular computer
capable of competing with electronic computers on classical computational problems in not war-
ranted.”^114 Of course, such breakthroughs may yet occur, and this possibility warrants some level of
continued research.
8.4.1.4 Future Directions
While it was DNA’s resemblance to the tape of a Turing machine that inspired Adleman to investi-
gate the possibility, this model has not yet been pursued experimentally. Nor is it likely that it would
have practical computing utility—a Turing machine is extraordinarily slow even executing simple
algorithms.
A very different approach would involve single molecules of DNA (or RNA or another biomolecule)
acting as the memory of a single process, while enzymes performed the computation by splicing and
copying sequences of bases. Although this has been discussed theoretically, it has not yet been shown in
an experiment. This model would be best used for massively parallel applications, since the individual
operations on DNA are still quite slow compared to electronic components, but it would offer massive
improvements of density and energy efficiency over traditional computers.
In a slightly different approach, enzymes that operate on DNA sequences are used as logic gates,
such as XOR, AND, or NOT. DNA strands are data, and the enzymes, by reacting to the presence of
certain sequences, modify the DNA or generate new strands. Thus, using fairly traditional digital logic
design techniques, assemblies of logic gates can be constructed. The resulting circuits will operate in
exactly the same manner as traditional silicon electronic-based circuits, but at the energy efficiency and
size of molecules.^115
Even if it turns out that biomolecular computation is a dead end, the research that went into it will
not be for naught: the laboratory techniques, enabling technologies, and deeper understanding of
(^114) R.S. Braich, C. Johnson, P.W.K. Rothemund, D. Hwang, N. Chelyapov, and L.M. Adleman, “Solution of a 20-Variable 3-SAT
Problem on a DNA Computer,” Science 296(5567):499-502, 2002.
(^115) M.N. Stojanovic, T.E. Mitchell, and D. Stefanovic, “Deoxyribozyme-based Logic Gates,” Journal of the American Chemical
Society 124(14):3555-3561, 2002.