A multinational European project has begun work on a biologically inspired, “wet” computer designed to mimic living brain functions through chemical assembly processes and pharmaceutical manufacturing techniques. The Neuneu project, for which the EU’s Future and Emerging Technologies (FET) Proactive Initiative will provide 1.8 million euros, will exploit several properties of chemical systems for their computing power.
“This is the first step towards real-life construction of an artificial chemical brain with well-defined architecture of connections between artificial neurons,” said professor Andy Adamatzky at the University of the West of England (UWE). “It will be a massive parallel computer made of lipid bubbles.”
The research will focus on building small networks and simulating large-scale networks of chemical microprocessors that oscillate using Belousov–Zhabotinskii (BZ) reactions. BZ bubbles are seen as a rough physical imitation of neurons. Both can be excited and go into a refractory state, and both support excitation waves traveling between elements.
Although the lipid-computing technique is an important step, it’s only a rough approximation to a brain. In a 2D plane, each lipid sphere can connect with five to seven neighbors, and the connections are only local. This pales in comparison with the 8,000 connections arriving at one neuron.
The project has three objectives. The first is to engineer lipid-coated water droplets, which contain the chemical medium. The droplets can be interconnected so that waves of excitation can flow between them. The second is to develop information processing architectures based on the physical and chemical properties demonstrated by the droplets. Third is to explore the limitations of these architectures.
The University of Southampton will refine the process of droplet construction. The Polish Academy of Science will model the oscillations of the BZ reactions from various chemical compounds. The UWE will develop simulations for modeling large networks of interconnected droplets. The University of Jena will translate the properties demonstrated in a lab to these large-scale models.
How It Works
The project makes use of stable cells, which are chemical bubbles coated in a fat-based membrane that forms spontaneously and uses chemistry to accomplish signal processing.
A biological cell has a bilevel membrane constructed from an array of lipid-molecule matched pairs that are sandwiched together, separating a watery medium inside and outside the cell. Each matched pair consists of a hydrophilic head that’s attracted to water and a hydrophobic tail that’s repelled by water. The hydrophobic molecules bond together and chemically attract other pairs to surround the cell.
Neuneu will employ a variation of this type of cell. A monolevel membrane surrounds each lipid droplet; it has water on the inside but only oil outside. A bilevel membrane is formed at the contact point where two of these bubbles are pushed together. The University of Southampton researchers are proposing a molecule of alpha–hemolysin to rip a tunnel between the two droplets, thus forming a channel between them.
Exciting the chemicals inside the droplets expends their chemical energy, causing them to enter a refractory period, during which they must recharge. An external excitation can then trigger another chemical reaction, much like a broken clock that ticks only when it’s jostled, then goes dormant until it’s can excited again.
The project will find ways to physically get the drops to touch and create connection points where they come together. In theory, researchers could explore different chemical mixtures for the droplets and different ways of making the channels. However, in the three-year project time frame, they will focus on proving the technology.
The theoretical research will address modeling the kinds of computing units the technology could support. Klaus-Peter Zauner, senior lecturer at the University of Southampton explained, “We want to find out if this is a technology that has real use or if it is too constrained.”
It’s difficult to make lab systems large enough to delineate these limitations, so the researchers will use simulation studies. Zauner noted, “We want to learn from small droplet networks with tens of droplets and extrapolate this to systems with 10,000 droplets. It’s not far beyond the scope of current technology, but it would be expensive to do in the lab.”
The lab research at this stage will refine the process for producing consistent quality droplets. The medical industry has developed the mass production of lipid-coated droplets as a technique for drug delivery.
BZ computation is a form of chemical computing used in molecular-computation research. It’s relatively easy to prepare, said Oliver Steinbock, associate professor of chemistry and biochemistry at Florida State University. In the simplest case, four chemicals (bromate, malonic acid, sulfuric acid, and a redox-catalyst) are mixed in water at room temperature. If stirred, the reaction solution can undergo long-lasting and striking color changes (say, red to blue) with typical oscillation periods of seconds to minutes.
External perturbations, such as submerging a wire in the solution, can trigger a long-lasting color-change cycle. This behavior is similar to the excitability waves that travel through neurons. The waves are like a domino-chain reaction, except the dominos can reset themselves after the refractory period. “The similarity between neurons and BZ systems has stimulated most of the chemical BZ computing ideas,” said Steinbock.
It’s likely to be some time before this type of research has practical application. “Don’t sell your PC yet,” Steinbock advised. “This is exploratory research that will not yield short-term applications although some highly specialized applications might be achievable. Nonetheless, the human brain is a convincing example that excitable networks can do remarkable things.”
The Molecular Computing Landscape
Zauner said the molecular computing landscape can be broken into three-broad areas: bulk-molecular materials, single molecules, and biomolecular computing.
Bulk-molecular materials, such as organic semiconductors, use soft-matter physics. The atoms are packed less densely and the structure is far less homogeneous than in solid-state semiconductors. Computation devices based on these materials uses technologies such as organic light emitting diodes. They can be more flexible and made at lower temperatures.
Single-molecule electronics uses novel construction techniques to build molecular wires, single-molecule diodes, and similar structures for smaller-scale devices. Both bulk- and single-molecule devices imitate conventional circuits with molecules.
Biomolecular computing uses the molecules in material-specific architectures. It’s the basis of cellular computation today. Researchers expect biomolecular arrays to enable architectures that are completely different from traditional logic circuits and will require entirely new programming techniques.
Neuneu is exploring this third type of molecular computing, which researchers have been exploring for some time. In the 1980s, researchers at the Biophysical Institute at Pushchino in Russia developed optical computers that leveraged biological components to store optical holographic information.
In 1989, Lothar Kuhnert reported processing images by using the BZ reaction with light-sensitive chemical waves. In the 1990s, Steinbock and his associates used BZ reactions to calculate the shortest paths in a labyrinth in a highly parallel fashion. They subsequently calculated Voronoi diagrams using BZ chemistry.
Zauner said the only commercial application of biomolecular computing so far is the FringeMaker-Plus, made by Munich Innovative Biomaterials. It uses the protein bacteriorhodopsin to create reusable media that store holographic image data for nondestructive testing.
The EU FET-Proactive Initiative is funding two other projects for molecular computing. The Bactocom project will use bacteria for computing. The Matchit project is building an infrastructure that uses DNA addressing to move chemicals. “There’s a lot of work going on,” Zauner noted, “and no one knows which techniques will ultimately work.”
Making It Practical
Neuneu might bring chemical computing from the concept stage to a practical demonstration. “There has been too much work on theory that was not tied closely to reality,” Zauner said, “so we were clear on tying the research to what can be demonstrated in the lab.”
The team has focused on techniques that have a reasonable chance of working well. For example, other researchers have made lipid-coated water droplets, but not for computational purposes, and BZ computation techniques are well developed.
The project aims to show how a new paradigm for building computers could be practical. Zauner sees traditional computer science locked into a somewhat narrow focus on certain forms of logical structures. “It’s almost as if you were to make a hot air balloon out of better and better materials, but never considered the possibility of an airplane,” Zauner said. “When we look at nature, it’s perfectly clear there are better computational techniques for many types of applications.”
Molecular techniques might be good for putting computers inside living cells — for example, to improve drug delivery so that robots could selectively detect and kill cancers. In the long term, Zauner expects that these kinds of computer will let us selectively create new molecules. “The big impact will be when molecular computing is used to make molecular materials that we cannot make today,” he said. “It will be more extreme than the impact of organic chemistry.”
Source | Computing Now