Over thousands of years, we’ve progressed from doing math with sticks and stones to mechanical calculators, and eventually to computers that perform billions of calculations using tiny bursts of electricity. Researchers have now found a rather unusual way to crunch numbers: multiplying bacteria.
Researchers at McGill University in Canada, working with an international team of scientists, have developed a biological computer that uses living bacteria to solve complex mathematical problems.
The experimental system allows millions of microscopic organisms to explore different solutions simultaneously, with the bacteria multiplying along the way to create even more computing agents. The researchers believe the technology could eventually tackle problems that would require enormous computing resources on conventional computers, potentially using far less energy.
Conventional electronic computers process instructions using transistors that switch electrical signals on and off. Although modern processors can execute many operations simultaneously, particularly with multicore CPUs and GPUs, certain complex problems still require examining an enormous number of possibilities. Parallel computing helps by distributing this work across multiple processors, allowing many possibilities to be explored at once.
The researchers have taken this parallel computing principle quite literally, replacing electronic processors with bacteria and computer circuits with microscopic channels. Their system, described in a study published in the journal Microsystems & Nanoengineering, is a microfluidic chip containing an interconnected network of tiny channels and junctions. Each channel represents a possible mathematical path, while the bacteria moving through them serve as the computing agents.
For their demonstration, the team used a classic mathematical puzzle called the Subset Sum Problem. Imagine being given the numbers 2, 5, 9, and 11, and asked which combinations add up to 16. You could test different combinations until you find 5 + 11, or 2 + 5 + 9. Easy enough with four numbers, but with 100 numbers, there are over a nonillion possible combinations to consider – a million trillion multiplied by another trillion.
The researchers translated this mathematical problem into a physical network of microscopic pathways. They first fabricated a precisely patterned silicon template, then used it to mould channels from a transparent, oxygen-permeable silicone polymer called polydimethylsiloxane (PDMS). They sealed the resulting network against a glass substrate, creating an enclosed maze through which the bacteria could swim.
Now, here's where the actual computing happens.
Specially designed junctions along the network act as logic gates. At a splitting junction, bacteria can take one of two routes, representing the decision to include or exclude a particular number from the sum. Other junctions allow bacteria to continue along their existing paths or merge into shared channels. The routes are arranged so that each valid journey through the network corresponds to a particular combination of numbers.
As thousands of bacteria swim through these junctions, they collectively explore different combinations in parallel. They don't actually know they're performing calculations, of course. Their movements simply follow the physical pathways encoded into the chip. At the other end, the channels terminate at exits representing possible sums. By observing which exits the bacteria reach and using automated image analysis to trace their routes, the researchers can determine which combinations satisfy the mathematical problem.
A particularly interesting thing about this computer is what happens while the bacteria are busy exploring.
Unlike conventional electronic processors, bacteria can naturally reproduce, effectively creating additional biological processors without anyone having to manufacture or introduce them. “Modern electronic computers make millions of operations per second using one chip,” says Dan Nicolau, the study's senior author and a professor of bioengineering. “With bacterial division, we have one million chips working at the same time. More chips mean you can resolve the problem faster, even if for each 'processor,' the number of operations per second is lower.”
To demonstrate this, the team experimented with two bacterial species: Escherichia coli and Vibrio natriegens. The former served primarily as a slower-growing reference, while the latter can double its population roughly every 9 to 12 minutes under favorable conditions. In one experiment lasting approximately 26 minutes, V. natriegens had enough time for two population-doubling cycles. The researchers also directly observed bacterial division inside the channels and measured more bacteria leaving the computing network than entering it.
The team successfully solved subset-sum problems involving up to seven numbers, with relatively low navigation error rates.
For example, E. coli incorrectly changed direction at just 1.3% of pass-through junctions, while V. natriegens did so at 0.3%. The researchers also tested larger, specially structured networks containing up to 200 repeated values of one. In the 100-value version, the network encoded 2,100 possible subsets, although these produced only 101 distinct sums. That makes it a useful demonstration of the network's physical scalability, rather than proof that the computer can solve arbitrary 100-number problems.
So, what exactly is the breakthrough here?
Biological computing and massively parallel computation are hardly new concepts. Researchers have previously demonstrated similar networks using molecular motors and other microscopic agents. The major difference is that the bacteria in this system can multiply while computing, allowing the number of available computing agents to grow alongside the problem being explored. The researchers describe their ambition as using biological multiplication to “defeat an exponential” increase in computational demands.
This doesn't mean the team has developed a bacterial equivalent of a modern CPU, or demonstrated a computer that outperforms today's electronic machines. The system is still an experimental proof of concept, and claims of superior speed and efficiency for larger problems are based partly on theoretical modeling. Practical challenges also include navigation errors, bacterial overcrowding, nutrient availability, and the considerable effort required to construct complex microfluidic networks.
Despite these limitations, the approach could eventually be useful for certain mathematical optimization problems, cryptography, molecular biology, and, potentially, aspects of artificial intelligence. “Our way is tailored for problems that electronic computers are not designed to solve, such as mapping the structure of proteins,” Nicolau says. “It was reported that AI figured this out, but this is only half-true, because it can only scan the available data. We propose this technology to solve problems, not estimate based on past information.”
Moving forward, the researchers propose several improvements, including programmable junctions that could change the network's logic, three-dimensional channels to reduce navigation errors, and engineered bacteria with better movement and reproduction characteristics. They will also need to demonstrate that the system can handle increasingly difficult, real-world problems without becoming overwhelmed by its own growing population.
Source: McGill University