DNA computing

DNA COMPUTING

INTRODUTION

Our topic is DNA Computing, an exciting new field where scientists are exploring the
computational power of biological molecules, in particular Deoxyribose Nucleic Acid, the basic
molecule of genetic expression. Whereas a conventional computer represents information on
silicon chips as a series of zeroes and ones, and manipulates the information by performing
mathematical computations on them, a DNA computer represents information as a pattern of
molecules in a strand of DNA .That information is manipulated by subjecting it to “precisely
designed chemical reactions that may mark the strand, lengthen it, or even destroy it” (Kierman).
Here my main purpose would be to look into the history and theory behind DNA computing and
its future potential.

A DNA computer is a molecular computer that works biochemically. It “computes” using enzymes
that react with DNA strands, causing chain reactions. The chain reactions act as a kind of
simultaneous computing or parallel processing, whereby many possible solutions to a given
problem can be presented simultaneously with the correct solution being one of the results.

The word “computer” is somewhat misleading in this context, as most people think of a computer
today as a machine that can generate word processing, produce spreadsheets, display graphics,
cruise the Internet and play MP3 files. However, at its core, it is a collection of electronic impulses
working across silicon-based circuitry. Electronic computers store information in binary form,
then reassemble and interpret that information in a meaningful way. A DNA computer has the
same basic ability to store information and compute solutions, though its methodology is different
in that it works off molecular automations, or preset reactions. Its greatest potential benefits
might lie in different areas that those of electronic computers.

For example, a DNA computer is a tiny liquid computer –- DNA in solution -- that could
conceivably do such things as monitor the blood in vitro. If a chemical imbalance were detected,
the DNA computer might synthesize the needed replacement and release it into the blood to
restore equilibrium. It might also eliminate unwanted chemicals by disassembling them at the

TheDirectData.com Page 1
DNA computing

molecular level, or monitor DNA for anomalies. This type of science is referred to as nanoscience,
or nanotechnology, and the DNA computer is essentially a nanocomputer.

The DNA computer is only in its early stages of development. Though rudimentary
nanocomputers perform computations, human interaction is still required to separate the correct
answer out by removing the DNA computer solution of all false answers. This is accomplished
through a series of chemical steps. However, experts are encouraged by the innate abilities of a
DNA computer and see a bright future.

Leonard Adleman, one of the pioneers of the DNA computer, reports that a single gram of dried
DNA is capable of storing the same amount of information as could fit on one trillion CDs. This,
along with the benefits of parallel processing and the negligible power required, guarantee that
the DNA computer, or nanocomputer, will continue to be refined and perfected. When molecular
computers become a reality, manipulation of matter at the level of DNA will lead to many
breakthroughs in all areas of science, industry, and medicine.

I. Solving Numerous Possibilities (NP) -Complete problems

A. Definitions:

1) A decision problem is in the class NP if there is no known algorithm for it that will execute in
polynomial time on a conventional computer, and it can be solved in polynomial time
using a nondeterministic computer.

2) A problem is intractable if no polynomial time algorithm can possibly be devised for it

(Miller). In other words, as the complexity of intractable problems increase, the time
required to solve those increases at an exponential rate (Adams).

3) A problem in NP is NP-complete if every other problem in NP can be expressed in terms of it

by means of a polynomial algorithm (Miller).

TheDirectData.com Page 2
DNA computing

If any problem in NP is shown to be intractable then all NP-complete problems are intractable.
However, if any NP-complete problem can be solved in polynomial time, all problems in
NP become tractable (Miller).

B. How is it done?

A test tube of DNA strands can be used to simulate a non-deterministic computer using the
techniques of molecular biology. Theoretically speaking, otherwise intractable problems can be
implemented in polynomial time on this DNA computer. One example is Adleman’s simulation of
the Traveling Salesman problem: known to be NP-complete, the problem had an execution time
that was linear to the number of cities (Miller). However, Adleman’s method of the exhaustive
search algorithm is able to solve problems in polynomial time only at the expense of exponential
increases in volume (Adams).

C. Implications to real life problems

Particularly in business planning and management science, many of the cost optimization
problems are NP-complete and currently solved using heuristic methods and other
approximations. These problems include scheduling, routing, and optimal use of raw materials,
and correspond to problems already solved, either theoretically or experimentally, using DNA
computation (Adams).

II. Cryptography

Currently, researchers have mapped a way in which DNA computers could be used to crack
messages encoded with Data Encryption Standard (DES) (Bergquist). Although this problem has
already been solved using conventional techniques in a much shorter time than proposed by the
DNA methods, the DNA models are much more flexible, potent, and cost effective (Adams).

III. Turing Machine

A number of theoretical models for creating a universal DNA computer have been developed, and
mathematical proofs have shown that some models of DNA computing are at least equivalent to a

TheDirectData.com Page 3
DNA computing

classical Turing machine (Adams). In year 2001, Ehud Shapiro’s group at the Weizmann Institute
in Israel built a programmable and autonomous computing machine using the DNA strands as
inputs and the transition rules. Shapiro’s machine was a “special case of the Turing Machine: a
two-state, two-symbol automaton” (Noble), but he is still hoping to develop a fully operational
molecular Turing Machine in the future (Parker).

Ehud Shapiro’s DNA computer

This model shows how a biological computer could work

It is being presented at the Fifth International Meeting on DNA-Based Computers at the

Massachusetts Institute of Technology.

In terms of the logic on which it operates, the prototype will behave in a similar way to molecules
inside a living cell, a "biomolecular machine".

The Turing machine uses the basic concepts of computing, reading and writing one bit of data and
performing an action depending upon a program. But although the Turing machine is a general-
purpose, universal, programmable computer and is a key to the theoretical foundations of
computer science, it has not been used in real applications.

Like a Turing machine, Professor Shapiro's mechanical device has a "rule molecule" designed so
that the processing of the molecule modifies another molecule in a predetermined way.

TheDirectData.com Page 4
DNA computing

To demonstrate the concept, Professor Shapiro has built a 30-centimetre-high plastic model of his
mechanical computer. He hopes that the advent of improved techniques for making and
assembling molecules will mean the day when his computer could be made is not far off.

If it were built from biological molecules it would measure about 25 millionths of a millimetre in
length, roughly the size of a cell component called Ribosome.

IV. Patterning polysilicon directly without lithography

According to Adleman, by laying a grid of DNA precisely on top of silicon substrates, very dense
patterns of nanocrystals can seed polysilicon circuits fabricated atop them. The accuracy of
DNA enables circuits to be built on a chip, molecule by molecule, instead of with bulk
deposition and lithography. Those substrates would pattern themselves onto flat sheets,
which could then be used to direct the placement of polysilicon materials atop them,
making up the actual circuit elements (Johnson).

Anchoring DNA to a substrate may liberate its computing potential. Scientists have made a small
but vital step forward in the quest to harness the vast computing potential of DNA.

TheDirectData.com Page 5
DNA computing

THEORY

DNA computing has certainly generated a lot of excitement, but there is a reason that it has not
yet replaced silicon as the primary platform for computing. Nearly 50 years have been spent
developing techniques and tools for programming traditional computers, and little of that effort
can be easily transferred to DNA computing. Traditional computer programs run on processors
specifically designed to run many different kinds of programs. DNA, however, has a very different
purpose: guiding life. Since it is not yet possible to adapt the tools of DNA to our purposes (almost
all useful biological molecules have been found in nature, rather than designed by humans), we
must instead adapt our programs to the tools available. Since the computational capabilities of
DNA are far more restricted than those of a modern silicon processor, we have to encode our
problems and algorithms in very unnatural ways. It is not entirely unlike translating an algorithm
to run on a minimal Turing machine.

It is interesting to compare DNA computing to the Turing Machine. Two things are needed to
build a universal computer: Storage and Processing. In the case of the Turing Machine, the
storage takes the form of the tape. In DNA computing, the DNA itself is storage medium, one of
the densest known. The processing of the Turing Machine is done by what Turing called the
"Finite Control," or what we might think of as the state machine. The processing of DNA
computing is a bit more complicated, and explained in the demo.

So if DNA computing takes so much effort, why is it such a hot topic? The immense power of DNA
can be summarized with two words: massive parallelism. It is generally believed that so-called
"NP" problems, like the Hamiltonian Path Problem, are fundamentally exponential. That is to
say, no matter what algorithm is used, the amount of computation required to solve a given
problem will be proportional related to an exponential function of the size of that problem. "An
exponential amount of computation" is normally interpreted to mean that a given computer will
take an exponential amount of time to solve, it. However, some algorithms can be "parallelized,"
which means that the computation is divided up between multiple independent computers. If a
parallelizable algorithm is used, then instead solving the problem with an exponential amount of
time on one computer, one can solve it using a fixed amount of time on an exponential number of
computers. This is called parallel computing and is what gives DNA computing its power. While

TheDirectData.com Page 6
DNA computing

an exponential amount of time on a conventional computer can add up fast, an exponential

number of "DNA computers" really just means adding more DNA to the beaker. For large
enough problems, the amount of DNA required can get unwieldy, the theory is that for some
problems, properly encoded, it will be easier to get the amount of DNA required to solve the
problem than the amount of computer time that would be required with traditional computers. It
is interesting to note that, according to Adleman, the most tedious part of the process is the
filtration by magnetic bead. In the Hamiltonian Path Problem, this step needs to be done once for
every node, except the first and last. Therefore, while the heavy computation is done in constant
time by an exponential amount of DNA, the extraction of the answer still takes a linear amount of
time, not constant.

So when will DNA computers replace silicon? I think never. While DNA is capable of amazing
feats of parallelism that have definite applications in supercomputing, the versatility, availability,
durability, quick responsiveness, and ease of interface of silicon far outweighs the benefits of DNA
computing in a desktop, server, or embedded environment. However the information density of
DNA does make it an attractive medium for mass storage.

The Genesis

The concept of DNA computing was introduced in 1994 by USC professor, Leonard Aldeman, in
the November 1994 Science article, Molecular Computations of Solutions to Combinatorial
Problems. Adleman showed that DNA could be used to store data and even perform computations
in a massively parallel fashion. Using the four bases of DNA (adenine, thymine, cytosine, and
guanine), Adleman encoded a classic “hard” problem (one that exhibits exponential growth with
each additional input parameter) known as the Travel Salesman Problem into strands of DNA
and utilized biological properties of DNA to find the answer. Adleman stumbled upon the idea of
DNA computing when he noticed how DNA replication was remarkably similar to an early
theoretical computer developed in the 1930’s by Alan Turing. During replication, DNA

TheDirectData.com Page 7
DNA computing

polymerase slides along a single DNA strand, reading each base and writing its complement on the
new strand, while in one version of the Turing Machine, a mechanism moved along a pair tapes,
reading instructions from an “input tape” and writing out the result on the “output tape”.
Interestingly, Alan Turing’s simple machine was proven to have the same computing capability as
any modern computer. Adleman now began to wonder: if Turing’s simple machine has such great
computational ability, would similarly operating DNA also have the ability to do computations? It
did, as Adleman’s first experiment proved. Although his experiment involved large amounts of
slow, manual labor to separate out the correct answers, included a high chance of error, and was
unscalable for larger problems, DNA computing promised immensely high density storage,
unparalleled energy efficiency, and a level of parallelism unknown to digital computers. A new
field was born.

First Steps

Adleman’s experiment sparked huge interest in the community and it wasn’t long before other
academics began exploring the other applications of DNA computing. Richard Lipton, a Princeton
professor described a method for solving the famous NP-complete satisfiability problem only

TheDirectData.com Page 8
DNA computing

months after Adleman’s discovery. In a method similar to Adleman’s, Lipton describes how all
possible combinations of values in an SAT problem can be described in a graph (show fig.1),
which could then be translated into DNA in the exact same manner as Adleman’s Traveling
Salesman example. According to Lipton, one could theoretically deduce the answer from this pool
by using a series of biological steps (involving pouring mixtures of specific strands around) whose
complexity was directly proportional to the number of statements in the SAT problem. In other
words, Lipton’s algorithm for solving the SAT problem was a linear one, a feat yet to be achieved
using modern day computers.

Although Lipton never published any actual test results during this time, he and his team
contributed greatly to the field with their theoretical papers on molecular computing. Lipton and
two Princeton graduate students, Dan Boneh (now at Stanford) and Christopher Dunworth,
published a paper on more efficient algorithms for solving general boolean circuits and other
optimization problems, also describing how to find the optimal setting for boolean networks for
which no complete solution existed. The following year (1996), the three published a well known
paper where they described a “molecular program” that could be used to break the Data
Encryption Standard (DES), a widely used cipher system approved by the US government. In the
paper, they claimed that their algorithm, given one plain-text/cipher-text pairing as input, could
determine the encryption key in approximately four months. Furthermore, the paper claimed that
if using a pair where the attacker specified the plain-text to be encoded, their algorithm could
theoretically recover the key about one day’s time.

Further Development

The Maximum Clique Problem

In 1997, a team of researchers from the NEC Research Institute finally showed some new
experimental results for solving NP-complete problems using molecular computing. Unlike the
vast majority of papers written since Adleman’s first crude experiment, this article was not purely
theoretical, as it described an actual experiment where researchers solved the maximum clique
problem given a six vertex graph. The NEC team demonstrated a new technique used in the
construction of the initial dataset which was considerable improvement over Adleman's. Instead

TheDirectData.com Page 9
DNA computing

of constructing each possible node and node-to-node connection individually as Adleman did, the
researchers only made one DNA strands for each node, and then employed a method (exploiting
the simple network structure of cliques) which would automatically generate all possible paths
between the nodes using a method called “thermal cycling.” However, removing illegal choices
(because not all nodes in the graph are interconnected) still involved manual splicing. Once the
data set was created, finding the maximum clique simply required using get electrophoresis to find
the longest clique. However, because the number of "unconnections" between nodes grew
exponentially just like the number of connections did, using manual removal of illegal
configurations made this algorithm impractical on a larger scale as well, even if we ignored the
many errors inherent in DNA processing.

RNA Solves the "Knights Problem"

As scientists around the world continued searching for more algorithms to classic problems for
DNA, a team of scientists from Princeton demonstrated a way of using RNA as a molecule for
computation in an experiment which solved the “Knights Problem” (how to place a number of
knights on a chessboard so that none of them are attacking each other) on a 3x3 board. Building
on the NEC team’s method of automatically building the dataset, the researchers created 18
different strands of DNA, one to represent each board positions with and without a knight. (9
squares x 2 possible choices/square = 18 different combinations). Using these 18 combinations, the
researchers created an RNA library containing the 512 (2^9) board configurations. The
researchers then described the problem as a SAT problem by assigning each square a variable
and setting them up in a series of logical expressions; this can be done for any “chess” problem.
The next step, where the incorrect strands are removed is where this experiment shines. Whereas
previous experiments required all kinds of manual cutting, splicing, mixing, and complex
screening to recover the correct answers in various test tubes, RNA could be easily digested using
an enzyme called RNase H. The researchers could simply digest incorrect RNA strands inside the
test tube using this enzyme. By using a series of branching, digesting, and then recombining, the
researchers could “execute” the entire Boolean statement, resulting in the possible board
configurations. In a random sample of 43 results, there was only one error which was attributed to
RNA mutation. The researchers concluded that their method of “in place” RNA digestion was
quite robust and could be applied to problems involving up to 2^50 RNA molecules.
TheDirectData.com Page 10
DNA computing

Hair pinning?

The trend in DNA computing was naturally toward more automation and more generalization. In
2000, an interesting concept was introduced: utilizing the biological structure of DNA to increase
automation. Previous experiments required large amounts of “old-fashioned, shake-the-test-tube
lab work for each step in the calculation” (Cho). The new method takes advantage of DNA’s
property of twisting itself into knots, hence the term "hairpin". Instead of manually cutting apart
strands using various enzymes, or even using enzymes to digest RNA to generate the all the
possibilities, this team designed strands that would automatically “fold over” when lowered below
a specific temperature. This allowed larger datasets because of the reduced manual labor, but it
also greatly increased erroneous results and required searching over many redundant solutions
because of the repeated complementary sequences required for the folding to be possible. Major
researchers in the area have expressed interest in this technique, but caution that they must
“reduce the large proportion of errors” (Landweber) for this method to be viable.

Recent Development

Bigger is better

The largest problem yet solved by a DNA computer was a 20-variable 3-SAT problem run by
Adleman and others in 2002. Using a similar encoding scheme to the one Lipton proposed
in his 1995 article on solving SAT, the team of researchers performed an exhaustive search
of over 1 million possibilities; a dataset eleven times bigger than the one used in the
Knights Problem (2^9 vs 2^20). To narrow down the massive data set, the researchers built
a "computer" consisting of an electrophoresis box containg a "hot" and a "cold" chamber.
The current dataset is contained in the hot chamber while the "clause" (describing one
truth statement) is in the cold chamber. Theoretically, during electrophoresis, the strands
that satisfy the clause in the cold chamber will migrate across and be "captured" in a
special layer of gel. These strands then become the "dataset" for the next clause and the
next clause is then "executed." Doing this for all twenty clauses should result in the results
that satisfy the entire statement. It did, as the researchers were left with the unique result
to the 20 clause problem.
TheDirectData.com Page 11
DNA computing

Other avenues in modern day research involve finding more efficient ways of performing the “test
tube” chemical operations used by Adleman and others to extract the correct solution strand. A
group of researchers at the University of Wisconsin are working on the design for a “high density,
chemically complex surfaces that can replace the test tube/magnetic bead methodology and
simplify the task of performing the thousands of successive separations by hybridization that will
be required in any DNA computing scheme.”

The first "real" DNA Computer?

Last year (2002), a group of Japanese researchers at Olympus Optical Co. announced that they
had created the “world's first DNA computer for gene analysis.” Built on the work done by
University of Tokyo Professor, Akira Suyama, Olympus claims to have built the first fully
automated DNA computer boasting huge computational capacity, low energy consumption, and
massively parallel processing capabilities. An important feature of the system is the electric
computer, which handles the lengthy process and complicated process of creating the complete set
of answers and then using DNA manipulation to remove incorrect answers. The computer also
increases accuracy as it cuts out human error. Olympus also introduced many important
technological advances. Most notably, artifical DNA that behaves in the same way as real DNA,
but hybridzes without interfering with eachother (in other words. virtually no mutations and no
error). The other notable technology is a technology called Magtration(R), which facilitates
precise seperation of specific DNA strands using magnetic microparticles. The creator says the
computer can be used to solve many different types of problems (unlike previous experiments
which were designed to solve a single instance of a problem) by simply changing the “source
program” of the electronic computer that controls sample handling and processing. Although
experts are cautious of calling this machine a real computer, it's a huge step in DNA computing.
Not only is it the first practical application of DNA computing in the commercial sphere, it leads
the way in what some consider the future of the field, namely integrating DNA's massive
parallelism with "traditional" digital computers to solve certain large/hard problems that are very
difficult (or impossible) to solve using digital computers alone.

APPLICATIONS:

TheDirectData.com Page 12
DNA computing

Here in two of the applications have been discussed with a brief demo of their working:

a) HAMILTONS PATH FINDING PROBLEM. b) THE MAXIMAL CLIQUE PROBLEM.

HAMILTONS PATH FINDING PROBLEM:

The Hamiltonian Path Problem was the first application of DNA computing. Find a path going
from the start to the end, passing through each other node exactly once. Can you do it in 54
seconds? 53. 52...

TheDirectData.com Page 13
DNA computing

Each city and each connecting flight is encoded as a DNA fragment. Each city can be thought of as
a pair of codes: a "landing runway" and a "takeoff runway." Each "runway" must have a distinct
code. Note that the "LA to Detroit" flight complements the "takeoff runway" of LA and the
"landing runway" of Detroit.

TheDirectData.com Page 14
DNA computing

The LA and Detroit fragments both bind to the connection fragment. An enzyme called ligase
(which fixes broken DNA strands in healthy cells) "welds" the fragments together. This process
repeats to form large legal paths. Somewhere in there is the solution.

TheDirectData.com Page 15
DNA computing

A protien called DNA polymerase attatches to a "primer" coded specifically for Los Angeles. It
then proceeds to produce the entire complement strand from ambient unnattached base-pairs.
This process "amplifies" or replicates the DNA strands.

One primer attaches to the starting city, and the other attaches to the complement of the finishing
city. This way, only a path with the proper start and end points is replicated both ways.

TheDirectData.com Page 16
DNA computing

This diagram shows how selective PCR (polymerase chain reaction) extracts segments with the
proper start and ending points from the zillions of strands of DNA.

TheDirectData.com Page 17
DNA computing

The DNA is forced through a gel matrix. Long DNA strands encounter more resistance than
smaller ones, which can simply slip through.

TheDirectData.com Page 18
DNA computing

By applying a current to the gel block, the negatively charged DNA molecules are forced through
the gel, and smaller ones get out ahead. They form bands that can be detected with ultra-violet
light. Fragments of the same length cluster together, and the band corresponding to the proper
path-length can simply be cut out. This technique is called "gel electrophoresis."

TheDirectData.com Page 19
DNA computing

An iron bead is attatched to a fragment complementary to the second city. A magnetic field is then
used to pull out all of the DNA fragments containing this city code. Those not containing it are
discarded. This process is repeated for all other cities in the path. At the end, each remaining
fragment must contain one copy of the code for each city. Since each of all the connections formed
were legal (because they were facilitated by "flight" fragments), each fragment that passed the
PCR test starts and ends at the right place, each fragment that passed the gel electrophoresis test
has the right number of cities, and each fragment that passed the magnetic ball test has one of
each cities, the remaining paths must be Hamiltonian!

The Maximal Clique Problem

Although many theoretical papers have been published on DNA computing since Adleman’s first
crude demonstration in 1994, it would be over two years, in 1997, until another NP complete
problem would be solved. This problem was the Maximal Clique Problem: given a group of
vertices some of which have edges in between them, the maximal clique is the largest subset of
vertices in which each point is directly connected to every other vertex in the subset. Every time a
new point is added, the number of total cliques that must be searched at least doubles; hence we
have an exponentially growing problem. Once again, researches sought to take advantage of
DNA’s high level parallelism which would essentially allow all the possible paths to be calculated
simultaneously. After all the possible cliques were constructed, the scientists would simply need to
fish out the largest clique. However, like many DNA experiments, each possible “choice” needed a
unique DNA strand. This is a problem because, as noted early, the number of cliques grows

TheDirectData.com Page 20
DNA computing

exponentially. In Adleman’s experiment, each city and every connecting path had to be
“hardcoded,” in other words, specially made to order in a biology lab. In a problem with more
cities, it would be virtually impossible to manually create every possible path. Ouyang and
company designed an algorithm that would only require the manual creation a linearly growing
number of DNA strands, 2N to be exact, where ‘N’ is the number of vertices in the graph.

The six-node graph for this problem

The maximum clique size is 4, and the maximum clique contains the nodes 2,3,4,5.

Their algorithm went like this. Each possible clique was represented by a binary number of N bits
where each bit in the number represented a particular vertex. If a certain bit held a ‘1’, the
corresponding vertex was in the clique, if it was a ‘0’, it wasn’t part of the clique. For example, a
clique containing the first four nodes would be “001111”.

Then, they removed those cliques that contained illegal connections; because not every node was
interconnected in the original graph (otherwise the entire graph would be one big clique). The
figure on the right shows the illegal paths. For example, any strand with a "1" in the "0" and "2"
spot could not exist (aka. "xxx1x1" is an illegal configuration). The resulting data pool held all
possible cliques of the graph and the string with the largest number of ones would be the
maximum clique.

TheDirectData.com Page 21
DNA computing

To create the complete data pool, the scientists utilized DNA pairing (each nucleic acid has a
complement, so if two strands contain a series of complementary DNA, they will stick to
eachother). The scientists arranged
the cliques so each vertex is
separated by a string of
“connection” DNA which we’ll call
P. Thus, a clique for the six node
graph would be represented as
P0V0 P1V1 P2V2 P3V3
P4V4P5V5 P6, where V is each
vertex. The trick is that each Pi of
adjacent vertices are
complementary strands of DNA, so
scientists need only create strands
of DNA for individual nodes (aka.
PoV0 P1, P1V1 P2 etc.) And they
will naturally bond together. Thus the researchers only needed to create two sequences of DNA for
each node, one to represent the node if it was in the clique (a ‘1’) and the other to represent if it

TheDirectData.com Page 22
DNA computing

wasn't in that clique (a ‘0’), and the complementary ‘P’ strands would bind them together to form
6 vertex long strands representing all possible cliques. In case you were curious, the naming
standard used was no nucleic acids if the vertex was a "1" and a predefined sequence of 10 acids if
the vertex was a '0'. Thus it would easy to determine the size of the maximum clique simply by
looking at the lengths of the strands. A method called “thermal cycling” was used to recursively
construct all possible strands, and the complete/correct strands (those that began with V0 and
ended with V5) were amplified.

The next step, removing cliques that cannot exist for the graph is a bit more difficult and requires
a large amount of manual labor. Although I will not go into the biology, the process involves
cutting strands at specific places using enzymes, dividing the data into separate test tubes, and
then using sequential restriction operations to eliminate the strands containing the illegal
connection; a process which must be repeated for every non-connection in a graph. Thus, even
with a faster construction method, this experiment is not scalable. Finding the final answer is easy;
gel electrophoresis will reveal to us the shortest DNA strand, which corresponds to the largest
clique.

Future Development

a. Nanotechnology

“Current biomolecular computing technology is still far from overtaking the silicon chip.
However, DNA computing seems to be the first example of true nanotechnology, forging a link
between computational science and life science. Solutions take multidisciplinary teams employing
molecular biologists, mathematicians, computer scientists, biochemists, and material engineers”
(Ellison).

Moreover, if researchers are able to control the molecular devices, including DNA, inside every
cell, they will be able to engineer devices more complicated and more efficient than current
microelectromechanical systems. For example, DNA computers could be used to “time-release
medications, bolster organ function, or provide medical feedback” (Srivastava).

TheDirectData.com Page 23
DNA computing

b. Change in the direction of the current progress

Other than developing a DNA computer to perform computer operations, there can also be
applications making use of the "natural" capabilities of DNA, including informational storage
abilities and interacting with existing and emerging biotechnology. One example would be using
the DNA technology inside cells (Amos).
Further potential applications might make use of the error rates and instability of DNA based
computation methods as a means of simulating and predicting the emergent behavior of complex
systems. This could pertain to weather forecasting, economics, and lead to more a scientific
analysis of social science and the humanities (Adams).

c. Complement (but not replace) today’s computers

DNA computers can specialize in large computational problems in which the number of possible
answers is enormous (Bergquist).

d. Implications to Biology, Chemistry, and Medicine

It is essential for the progress in DNA computing to have high levels of collaboration between
different academic disciplines such as computer science, mathematics, natural science, and
engineering. Other than continuing the development of a practical DNA computer, this
collaboration can contribute to an increased understanding of DNA and other biological
mechanisms (Adams).

e. DNA's Role in Computer Science

DNA has the potential of being a natural storage medium and a tool of computation. Despite the
high error rates encountered in DNA computing, in nature DNA has little understood but resilient
mechanisms for maintaining data integrity. An increased understanding about the limitations of

TheDirectData.com Page 24
DNA computing

computation with come with the advent of DNA computational paradigms directed towards
improved methods of solving NP-complete problems. Additionally, “DNA based computers may
postpone certain expected thermodynamic obstacles to computation as well as exploring the
limitations of Turing machines and questioning theories of computation based on electronic and
mechanical models”

f. Biomolecular computation

For example, enabling a computing system to read and decode natural DNA directly. Such a
computer also might be able to perform DNA fingerprinting—matching a sample of DNA, such as
that in blood found at a crime scene, with the person from whom it came. The DNA computer
might also be a cost-effective way to decode the genetic material of humans and other living
things. This would eliminate the time-consuming task of translating DNA to store electronically
and create wet-data-bases of DNA for research purpose.

g.Industry

“While most research is taking place at universities, some companies are probing the potential of
DNA computers. NEC Corp.'s Research Institute in Princeton, N.J., for instance, has several
scientists working on DNA computing. Hewlett-Packard Co., in Palo Alto, Calif., is keeping tabs
on 6 to 10 major projects” (Ellison).

Conclusion

DNA computing is so exciting because of the collaboration of chemists, biologists, mathematicians,

and computer scientists to understand and simulate fundamental biological processes and
algorithms taking place within cells. Although DNA computers might not replace conventional
computers in the near future, they still have endless potentials for other applications.}

So will DNA ever be used to solve a traveling salesman problem with a higher number of cities
than can be done with traditional computers? Well, considering that the record is a whopping

TheDirectData.com Page 25
DNA computing

13,509 cities, it certainly will not be done with the procedure described above. It took this group
only three months, using three Digital AlphaServer 4100s (a total of 12 processors) and a cluster of
32 Pentium-II PCs. The solution was possible not because of brute force computing power, but
because they used some very efficient branching rules. This first demonstration of DNA
computing used a rather unsophisticated algorithm, but as the formalism of DNA computing
becomes refined, new algorithms perhaps will one day allow DNA to overtake conventional
computation and set a new record.

On the side of the "hardware" (or should I say "wetware"), improvements in biotechnology are
happening at a rate similar to the advances made in the semiconductor industry. For instance,
look at sequencing; what once took a graduate student 5 years to do for a Ph.D thesis takes Celera
just one day. With the amount of government funded research dollars flowing into genetic-related
R&D and with the large potential payoffs from the lucrative pharmaceutical and medical-related
markets, this isn't surprising. Just look at the number of advances in DNA-related technology that
happened in the last five years. Today we have not one but several companies making "DNA
chips," where DNA strands are attached to a silicon substrate in large arrays (for example
Affymetrix’s genechip). Production technology of MEMS is advancing rapidly, allowing for novel
integrated small scale DNA processing devices. The Human Genome Project is producing rapid
innovations in sequencing technology. The future of DNA manipulation is speed, automation, and
miniaturization.

And of course we are talking about DNA here, the genetic code of life itself. It certainly has been
the molecule of this century and most likely the next one. Considering all the attention that DNA
has garnered, it isn’t too hard to imagine that one day we might have the tools and talent to
produce a small integrated desktop machine that uses DNA, or a DNA-like biopolymer, as a
computing substrate along with set of designer enzymes. Perhaps it won’t be used to play Quake
IV or surf the web -- things that traditional computers are good at -- but it certainly might be used
in the study of logic, encryption, genetic programming and algorithms, automata, language
systems, and lots of other interesting things that haven't even been invented yet.

TheDirectData.com Page 26
DNA computing

TheDirectData.com Page 27