While meteorological institutes produce global weather forecasts, many different data flows to these forecasts throughout the process. Weather researchers use computers with many different processors to better cope with the amount of data. Each of these processors, called a “parallel computer”, simultaneously evaluates data from a particular region of the world.
In addition to the arithmetic units that perform the actual operations, such a computer also needs many processors just to distribute the operations. Simply put, they assign tasks to arithmetic units and somehow combine partial results into a logical aggregate result. In the weather forecast example, the controlling processors individually control calculations from sub-areas of the Earth. They regulate the exchange of data and results among themselves. When we look at it, it is useless to predict how a high pressure area will behave if the influence of low pressure areas in the immediate vicinity is not taken into account.
Computing Power for the Control System
However, you cannot simply interconnect any number of processors to solve any number of complex problems. The more computers run in parallel, the more computing power is required for the control system. The effort required for control purposes often increases disproportionately. This means that there are only a few applications where parallel computing seems like a viable option.
The reason lies in the architecture of a digital computer. It was designed to run one computing step after another. The increasing computing power of computers in recent years has enabled digital computers to solve increasingly complex problems despite this obstacle. However, the further development of digital computers constantly faces physical and technical limits. Thus, to make parallel computation simpler, other research approaches to computation are of interest.
is winning.
In the shadow of quantum computers, which attract great attention with their promising research results in this field, many teams are also taking their steps with approaches in the field of biology.
What is a Bio-Computer?
The international research project Bio-4Comp has been dealing with networked biocomputers for several years. The idea is that biological agents are sent on a journey through a complex network of nanochannels that represent a mathematical problem. A number is then added or not added to the cumulative result, depending on which turn the agent takes at an intersection. The path the agent takes over the network corresponds to a possible outcome.
The advantage is that you can send many agents over the network at the same time and they can monitor all potential paths at the same time. Therefore, instead of properly calculating one solution after another like a traditional digital computer, the network-based biocomputer needs to do the calculations in parallel. of the Fraunhofer Institute for Electronic Nano Systems (ENAS) in Chemnitz. Danny Reuter is responsible for research work on the fabrication of networks and the scalability of technologies.
Dr. Reuter makes a comparison between crowd surfing at a rock concert and processes in the network: “Motor proteins move biological agents, which in our case are molecules derived in animal cells, just as music fans carry a singer through an audience.” So here the team is moving agents across the network. It uses the kinetic energy of motor proteins to
Thomas Blaudeck, Reuter’s colleague, also from Fraunhofer ENAS, hopes to have millions of agents in a network in the future to move from basic research to applied research: “Each agent is its own processor. Since moving in the nano network is much slower than the computing speed of a traditional digital computer. , we need a large number of intermediaries to take advantage of our advantages in practical applications.“
Viruses as Processors
These advantages are primarily related to parallelism and energy efficiency. These are exactly the areas where digital computers face challenges. Blaudeck sees potential applications of the network-based biocomputer in principle in all tasks of exponentially increasing complexity of possible combinations in each choice. “The advantage we have with biological approaches is material. Because it can reproduce itself under certain conditions.” says. At Bio4Comp, teams work with dead matter that has no life of its own.
However, molecules acting in the network as agents can split at intersections, for example, and thus perform two computational steps simultaneously. The first part adds the number represented by the intersection, the second part takes a different path and does not add the number.
However, other research projects are already working with live agents and sending viruses or bacteria through networks. Here, agents can simply multiply to increase the number of processors. This duplication is first and foremost the most necessary action. Because a kind of “bottleneck” is created at the entrance of the network. There, only a limited number of agents can enter the network at any one time. But the network branches further and grows larger with each pass.
Networks that truly allow for practical computations need a large number of intersections to represent a complex problem. Blaudeck explains, “Agent density, that is, the number of agents coming from a channel segment per unit time, gets smaller and smaller towards the exit. Then biology helps us with this problem.” he explains.
A Complement for Supercomputers
One day, bio-computers may also come to the fore in energy efficiency. According to Danny Reuter, these computers cannot replace the personal computers that sit under most people’s desks. “Our computers were designed to complement supercomputers. Any problem we want to solve with biocomputers can also be solved by supercomputers. But we hope one day to be faster and use a lot less energy to do the same calculations.” According to two Fraunhofer researchers, three to four orders of magnitude—less energy per calculation—is the goal of their project.
However, there are a few more hurdles to overcome on this path. “So far we have been able to demonstrate that the approach works to a reliable degree,” Reuter said. Right now our results are where quantum computers were three or four years ago and are still far from competing with supercomputers.” says. The crux of the matter is scaling, the main focus of Reuter and Blaudeck’s Fraunhofer team. “As we continue to grow our networks and send more agents, the space we need for a related issue is huge,” Reuter said. The error rate would also be very high.” he states. This is where the next construction site is seen.
For example, pseudo-tagging should improve the performance of computers. In this process, researchers mark molecules as they pass through the network so they can better read where they’re going later. Reuter reports that currently the agent is still viewed with a microscope as it passes through the network. “But we are working on electronic components that receive a signal when the agent passes by, or add some DNA to it at a certain point in the network, and then follow which path it takes.”
This will also facilitate detection at the exit of the network, which will be automated in the next step.
Not Especially Sustainable Yet
Also interchangeable junctions are missing in the project. So far, a nanonetwork represents only a single math problem. Thomas Blaudeck explains that the networked computer blurs the boundary between hardware and software: “In our case, software is represented in hardware by the exact arrangement of junctions.” A separate chip for each computation is something the researchers agree, yet not particularly sustainable. However, various computational problems can be represented and computed with a single chip if universally interchangeable intersections can be implemented.
While many questions remain unresolved, Reuter and Blaudeck are in an optimistic frame of mind. The biotechnology and manufacturing technologies required to produce nanochannels already exist. The challenge is both to bring together scientific disciplines with mathematics and computer science, and to develop sub-elements unfamiliar to classical microelectronics.
DNA to Calculate Square Roots
Meanwhile, research teams are tapping into other applications in biology. Computer scientist and molecular biologist Leonard Adleman conducted experiments with a programmable DNA in 1994. It then represented input values in DNA sequences where it reacted with each other in a test tube. With this, Adleman was able to perform simple mathematical calculations.
In 2019, another team was able to calculate the square root with such a DNA computer. Each DNA strand was assigned its own fluorescent color value. The new combinations of these color values after the experiment then corresponded to the result of the calculation.
The advantage of this approach is massive parallelization, as with the network-based biocomputer. DNA strands react with each other in all combinations at the same time in the test tube.
In theory, it’s particularly well suited for optimization problems. There are always several viable solutions to these problems. But one of them is the best, fastest and most economical. The best-known example is the traveling salesman problem. A trader should cover all cities in a list without visiting any of them twice. There are countless travel note options that come his way, but he naturally wants to take the shortest route to save kilometers.
Evaluation Methods Not Available
In the DNA computer, each city would receive its own DNA strand. They would all react to each other by duplicating a “nuthouse”, thus creating all conceivable paths at once. It would take years for a digital computer to do this calculation for a certain number of cities. If you now remove the longer pieces of DNA by targeted chemical reactions, theoretically the shorter path remains of all.
Here’s the trick: Practically, there are no suitable methods to evaluate the results after the reaction. It is not impossible for these procedures to be developed in the future and not so practical. DNA computers will be able to process related problems.
Dominik Heider, a professor of data science in Biomedicine at Philipps Marburg University, is still somewhat skeptical of DNA-based computers: “From an academic point of view, all this is pretty interesting. But I fear it will continue to be irrelevant in practice.” He says the reason for this is quite simple; Everything that DNA computers can do, quantum computers can do. Heider says it’s also much easier to deal with them. “For videotapes, as in the days of VHS and Betamax, only one of the two approaches will work, and I doubt it will be DNA computers.”
Binary Data Translated into DNA
However, Heider is by no means willing to give up DNA for computer science. In the MOSLA Research Project, he is working with colleagues from computer science, biology, physics and chemistry to store data in DNA. To do this, binary data from traditional digital computers, i.e. a long chain of zeros and ones, is translated into letters A, C, G, and T.
These letters represent the four bases. These letters mean the four bases that make up each DNA. Every genome of a living thing consists of an individual combination of these four bases. The translation can be easily transferred to real DNA, which can be stored for a long time in the laboratory and read again at any time. A digital computer can then convert the DNA data back to binary code and display the file digitally again.
However, Heider says there is a fair amount of computer science behind it in practice to ensure that no data is lost on the way: “There are sources of error during DNA synthesis, DNA storage during replication, and replication. During storage and sorting, our task is to develop fix codes that catch these errors.”
As with any storage, he said, there is a trade-off between storage density and cost: “It obviously costs more to store more data. Again, we cannot fit the code into an infinitely long piece of DNA. We need individual short pieces, and then always information on how to reassemble pieces correctly after sequencing.” But this meta information also takes up storage space.
Should Be Stored In A Cool, Dry And Dark Place
This DNA storage is still very expensive. “Until now, there was no need to produce such large amounts of DNA,” Heider said. says. Therefore, a process that is inexpensive enough to make the use of DNA storage feasible in practice still does not exist. However, Heider hopes that in ten years, with enough research, things could look radically different.
For some applications, DNA storage has many unique advantages: “- Our technology will be used primarily for long-term archiving. Data like historical documents, birth records, or long-term weather data that no longer changes are just perfect for DNA.” Once produced, storage requires almost no energy other than the operation of the refrigerator. That’s because DNA is easy to hide.” Heider said that; cold, dry and dark”. Error correction needs to compensate for individual mutations.
Most people will probably continue to store their vacation photos on hard drives, SSDs or the cloud. However, DNA storage may soon be the biological alternative for the large amounts of data in archives that no one needs to access regularly but are stored only for emergencies.
38782
