- Parallel computing can ‘fix’ the problem with conventional computers, that they are sequential.
- If operations could be performed in parallel, then combinatorial tasks could be performed quicker.
- This approach used biological filaments driven by motor proteins to solve a problem encoded in a network of channels.
The collaborative effort of researchers from Germany, US, Sweden, UK, Canada, and the Netherlands has resulted in the creation of a biological parallel computer. The research, recently published in Proceedings of the National Academy of Sciences (PNAS), has the potential to solve the problem with conventional computers — that they process only one computational task at a time.
“The limitation of conventional computers is that they usually perform one operation after the other, sequentially. That means whenever you have a problem of combinatorial nature you have to either run the problem in a trial and error way, or you have to try every possible solution,” says Stefan Diez, head of the research team at Technische Universität Dresden. “This means that the larger the problem, the larger the time the computer needs to run, and this increases in an exponential manner.”
The network solves the problem by using junctions to guide biological filaments driven by molecular motor proteins. Junctions cause the filaments to turn to the left or right depending on their angle. Each of the filaments do their own task in parallel to others, allowing them to independently traverse the network at a high number without relying on what the other filaments are doing.
Two biological systems were tested. Researchers in Germany experimented with microtubule filaments driven by kinesin motors, while Swedish researchers tested actin filaments driven by myosin motors. As far as the researchers are aware, this approach has never been attempted.
The method could be implemented with existing technologies, and is highly energy efficient, thus avoiding the overheating issues often encountered in electronic computers.
“The idea of encoding problems in these kind of networks has been around for a while,” says Diez. “To actually experimentally implement these ideas has taken us the last 5 years. The next milestone will be to demonstrate the ability to upscale the technology. This means the biological agents have to function reliably enough to work over longer times and larger distances, which we believe they can.”
This research demonstrates the principle on a very small scale; if the technology can be scaled up, then it would be possible to solve larger problems with it. But this doesn’t mean that biological parallel computing will overtake conventional computing in the near future.
“We are trying to put together a research team to get this off the ground, and we think within 3-5 years that we can set up a network that very efficiently calculates large problem sets," says Diez. “It won’t necessarily be faster than a conventional computer just yet, but ideally it will be faster than other alternative computing approaches.”