Can we use individual bio-molecules as computers? Researchers in the H2020 FET project Bio4Comp are working towards exactly that. Now they need your help with the next step.

Network-based biocomputation.

Imagine a high-performance computing cluster that runs on a few kilograms of biochemical fuel instead of megawatts of electrical power. This may sound like science fiction now, but within the Bio4Comp consortium, which is supported by the EU's Future and Emerging Technologies (FET) programme, we have demonstrated that, at least in principle, this is actually possible. Considering that in 2015 the world’s data centres consumed more energy than the entire UK, new technology that could greatly reduce the energy consumption of computation is in high demand. But how can bio-molecules perform mathematical calculations? The Bio4Comp team works with motor proteins, which transport cargo along protein filaments within the cells of our bodies:

An animation of a motor protein moving along a cytoskeletal filament in a cell.

Motor proteins have been optimized by more than a billion years of evolution to convert chemical energy directly into mechanical work. They are only a few tens of nanometers wide, highly energy efficient, and work in parallel. We take these motor proteins out of cells and put them into networks of channels that are created by the same technology that is used to fabricate computer chips. Because the motor proteins are attached to the floor of the channels, they transport their counterparts – filaments made of protein – along the channels and explore the network. The protein filaments can accordingly solve mathematical problems if the network is designed in such a way that finding a path to a particular exit solves the problem.

As a proof of concept, we designed a network that solves the subset sum problem, a special case of the "knapsack problem". The knapsack problem can be explained as follows: imagine your house is on fire and you can carry only one bag of items to safety. Which items should you pack so that the combination of items has the highest possible value? The difficulty is that you need to try all possible combinations of items in order to find the best fit. This is easy if you have only a handful of items in your house, but with an increasing number of items, the number of combinations you need to try increases exponentially. For example, for three items, you need to try 8 combinations (23), but for 20 items you already need to try more than a million combinations (220). Mathematically, the subset sum problem is very similar to the knapsack problem. It is defined as follows: given a set S of whole numbers, can a certain target number T be reached by summing up any combination of numbers from the set S?

We estimate that this type of network is approximately 5000 times more energy efficient than the most energy efficient electronic computer. This means that, instead of having to build data centres in Iceland where electricity is cheap and cooling efficient (half of the power consumed by conventional data centres is consumed for computing and the other half for cooling), we could build them anywhere and power them by a few kilograms of biochemical fuel instead of megawatts of power.

However, the networks we demonstrated so far are not as efficient as they could be. They store information in the position of the filament within the network. This was ideal for a proof of concept network because it is easy to explain and made the experiments simpler. Unfortunately, this means that the networks grow relatively large, which currently limits the applicability.

The next breakthrough in biocomputing would be the ability to store information directly on the filaments themselves. In this way, the networks would become much smaller. For example, we could encode the {2; 5; 9} problem with three junctions instead of 150 if we were able to store information on the filaments. This is where you come in: we are offering an award of € 5000  for the best idea on how to encode information into protein filaments. All scientists, students or professionals interested in the field of biocomputation can participate.[1] Please do not hesitate to send us even your crazier ideas – out-of-the-box thinking is crucial to solving this problem! The deadline for submission is August 15, 2018. All details for the award and our contact details in case of questions can be found on our homepage.


[1] Employees of the partner institutions of Bio4Comp (Lund University, Technische Universität Dresden, Linnaeus University, Molecular Sense Ltd, Bar Ilan University, Fraunhofer Gesellschaft) are not eligible for the award.