Virtual DNA replicates
Technology Research News
The key to life is self-replication. If
an organism can't keep its genetic material going by passing it on to
another generation, it fades from view rather quickly. Replication also
underpins evolution: throw in a little change variation over a large amount
of time, and you get many different organisms.
Self-replication is all around us, but it's not a simple process.
Artificial life researchers from the Canadian National Research
Council and the University of Waterloo have found a way to examine the
phenomenon more closely using a computer simulation of self-replicating
strings of symbols that work as a simplified sort of DNA.
The work promises to provide a better understanding of the nature
and origins of life. It also lays the groundwork for inexpensive and flexible
manufacturing processes that borrow from life's vast experience, including
the possibility of growing machines in vats of chemicals.
The simulation consists of T-shaped virtual objects that exist
in a two-dimensional virtual space, and are affected by several forces.
The objects "form chains and the chains replicate, much like DNA replicates,"
said Peter Turney, a senior research officer of at the Canadian National
The objects are like DNA's codons, said Turney. Codons are sequences
of three nucleotides in a string of genetic code; the portion of DNA that
makes up a gene is much longer. Genes provide a blueprint for making a
specific protein. Each codon specifies a particular amino acid needed
for that protein. The virtual objects assemble into patterns similar to
the way codons make up strands of DNA or RNA.
The researchers' simulation, named JohnnyVon after John von Neumann,
a founder of the field of computing who did some early work on self-replicating
cellular automata, consists of a virtual soup of two types of particles
the drift about in a simulated liquid.
The researchers programmed into the simulation several forces
that affect interactions among the particles, allowing them to make and
break bonds with each other.
The forces are attraction; repulsion; Brownian motion, or random
jostling by water molecules; viscosity, or the tendency of water to impede
motion; and momentum. The attraction and repulsion forces were inspired
by the electrostatic attraction and repulsion of atoms and molecules,
but are modeled more like the motion of a spring, said Turney. The objects
are also affected by spring dampening -- the oscillations slow down over
time because a certain amount of motion converts into heat.
The simulation takes place in a two-dimensional space rather than
the three dimensions of reality for practical reasons. It's "a trade-off
between computational complexity and realism," said Turney. "It is easier
to compute a two-dimensional model than a three-dimensional model," he
When the researchers seeded the simulation with a pattern of already-bonded
virtual objects, the pattern replicated itself -- the separate particles
arranged themselves into the same type of chain, according to Turney.
"In nature a seed pattern -- a particular string of DNA -- can replicate,"
Evolution requires that patterns -- in the form of DNA -- replicate
with characteristics that are inherited, that mutate, and that they do
so using a selection process that favors the replication of some patterns
over others based on inherited characteristics. The seed pattern results
show that JohnnyVon has the properties required for evolution, said Turney.
Achieving self-replication is the first step toward bringing life
processes to manufacturing, said Turney.
The same principles could eventually lead to real, nano-scale objects
self-assembling in vats of liquid in a manufacturing plant, he said. A
nanometer is one millionth of a millimeter, or the size of 10 hydrogen
atoms in a row.
"Imagine manufacturing an automobile by dropping a seed pattern
into a vat of nanobots suspended in liquid. Different seeds would yield
different automobiles or entirely different objects," said Turney.
Self-replication is only the first step toward that end, he added.
The specific patterns of DNA represent plans for building proteins, which
eventually assemble into cells and bodies. In JohnnyVon, however "seed
patterns can only replicate; they do not yet encode plans for building
things," said Turney.
Making the self-replicating simulation meant solving several challenges,
said Turney. One problem was finding the right balance between realism
and computational complexity, he said. The researchers had to make a model
that could plausibly be converted to a real physical system, but efficient
enough to run on a desktop computer, he said.
They also found balancing the different forces difficult, he said.
"It was... a challenge to tune the model to yield reliable, stable, orderly
self-replication and avoid spontaneous and chaotic behavior, such as excessive
mutation," he said.
Another difficulty was keeping the model purely local and distributed,
meaning no centralized control system was directing the process, he said.
This is important because local control systems are more lifelike, and
also more scalable and robust, or adaptable, than centralized control
systems, he added.
The work is a useful contribution to artificial life research,
said Jason Lohn, a computer scientist at NASA. "I've done and seen similar
work... the novelty [here is] the set of rules the authors have devised
that govern interactions," he said.
These rules are potentially useful "in our quest to understand
how much complexity is required for self-replication," said Lohn. They
also may help in advancing researchers' knowledge of self-assembly techniques,
The researchers' next step is to come up with patterns that encode
instructions for building things, rather than merely replicating themselves,
said Turney. "In technical terms, we currently have pure genotypes, e.g.
raw DNA, with no phenotypes e.g. protein, cells, bodies," he said. "The
next step is to add phenotypes."
It will be more than a decade before this type of work could be
ready for use in practical manufacturing processes, said Turney.
Turney's research colleagues were Arnold Smith of the Canadian
National Research Council, and Robert Ewaschuk of the University of Waterloo
in Canada. The research is scheduled to appear in the winter, 2003 issue
of the journal Artificial Life. The research was funded by The National
Research Council of Canada.
Timeline: > 10 years
TRN Categories: Artificial Life and Evolutionary Computing;
Story Type: News
Related Elements: Technical paper, "Self-Replicating Machines
in Continuous Space with Virtual Physics," Artificial Life, winter, 2003;
technical paper, "JohnnyVon: Self-Replicating Automata in Continuous Two-Dimensional
Space," arxiv.org/abs/cs.NE/0212010; JohnnyVon Java applet and movies
February 26/March 5, 2003
Handhelds gain space
Ceramic yields under
Virtual DNA replicates
catches the bus
bangs out plastic circuits
forms nano piston
could lower fuel-cell cost
drive tiny toolset
promises massive storage
bursts reveal hot topics
Research News Roundup
Research Watch blog
View from the High Ground Q&A
How It Works
News | Blog
Buy an ad link