DNA computing using single-molecule hybridization detection
Kristiane A. Schmidt
1
2
Christiaan V. Henkel
1
2
Grzegorz Rozenberg
0
1
Herman P. Spaink
1
2
0
Leiden Institute of Advanced Computer Science, Leiden University
, Niels Bohrweg 1, 2333 CA Leiden,
The Netherlands
1
Leiden Center for Natural Computing
2
Institute of Biology, Leiden University
, Wassenaarseweg 64, 2333 AL Leiden,
The Netherlands
DNA computing aims at using nucleic acids for computing. Since micromolar DNA solutions can act as billions of parallel nanoprocessors, DNA computers can in theory solve optimization problems that require vast search spaces. However, the actual parallelism currently being achieved is at least a hundred million-fold lower than the number of DNA molecules used. This is due to the quantity of DNA molecules of one species that is required to produce a detectable output to the computations. In order to miniaturize the computation and considerably reduce the amount of DNA needed, we have combined DNA computing with single-molecule detection. Reliable hybridization detection was achieved at the level of single DNA molecules with fluorescence crosscorrelation spectroscopy. To illustrate the use of this approach, we implemented a DNA-based computation and solved a 4-variable 4-clause instance of the computationally hard Satisfiability (SAT) problem.
-
Biomolecular computing studies the use of DNA or other
biomolecules for solving various computational problems
(13). Owing to the inherent logic of DNA hybridization
and the massive parallelism intrinsic to molecules, DNA
computers have the potential to extend the range of solvability for
computationally hard problems (2,3). Parallel search
algorithms have been employed in a number of experiments
for solving small-scale instances of such problems, e.g. the
Hamiltonian Path problem (1) and the Satisfiability (SAT)
problem (47). A second branch of DNA computing research
investigates the development of DNA-based nanodevices.
Examples are the DNA finite automata (8,9) and the
realization of logic gates in single deoxyribozymes (10).
The two areas of research are related, and while they may
both yield important applications, future molecular computers
that combine both approaches also hold considerable promise.
For example, instead of utilizing huge amounts of electronic
computer power to perform relatively simple analyses on vast
quantities of biochemical information, it might be possible to
construct a molecular computer, which efficiently processes
these data at the molecular level.
For the successful implementation of DNA-based
computations, the detection of output molecules is of prime
importance. Many of the currently available techniques for the
detection of DNA have been used in molecular computing:
gel electrophoresis with fluorescent or radiometric
visualization, fluorescent labelling and fluorescence resonance energy
transfer (FRET), mass spectrometry or surface-based
techniques. However, all these methods either detect DNA in bulk
quantities or destroy the output molecules. This severely limits
the size of the library to be searched: the largest parallel
computation reported filtered 220 different molecular species
(7), which is less than the number of molecules of one variety
necessary for the detection using gel electrophoresis (35 pg)
(11). For molecular automata, this detection limit imposes an
equally severe redundancy. Therefore, the application of more
sensitive detection technology may significantly enhance the
power of DNA computations.
Recent progress in optical detectors has enabled the efficient
detection of single molecules by fluorescence microscopy
(12). One of the most prominent single-molecule techniques
for biological research is fluorescence-correlation
spectroscopy (FCS) (13,14). FCS studies fluorescence fluctuations
caused by a single molecule diffusing in the focal detection
volume. Since binding of a small fluorescently labelled
molecule to a larger ligand results in the change in diffusion time,
FCS allows quantification of the interaction of biomolecules at
extremely low concentrations. Extension of the method to
dual-colour fluorescence cross-correlation spectroscopy (15)
circumvents the need for a mass difference between the
binding partners.
In this study, we report the detection of single molecules of
DNA performing a computation. Our procedure for
experimental implementation relies on the so-called blocking
algorithm (16), a parallel search algorithm which involves
direct inactivation of those molecules that are not a solution.
Fluorescence cross-correlation spectroscopy was employed to
detect hybridization between single DNA molecules. We have
benchmarked this technology on a small instance of the NP
complete SAT problem.
MATERIALS AND METHODS
The library for a 4-variable SAT problem (24 possible
solutions) was encoded by 16 different oligonucleotides of 36 bp
each (Table 1). They have the general structure:
50 start - a b c d - stop,
where start and stop are a leader and end sequence (...truncated)