How Can Digital Data Stored as DNA Be Manipulated ?

From CNRS (Centre National de la Recherche Scientifique)

Summary :

• Data can be encoded as DNA but are difficult to process thereafter.

• New method enables operations to be performed on DNA-encoded data directly, without having to first translate them into their electronic equivalent.

DNA can be used to reliably store a vast amount of digital data. However, retrieval or manipulation of specific data encoded in these molecules has hitherto been difficult.

Now, researchers from the CNRS, the ESPCI Paris-PSL and the University of Tokyo have pioneered the application of a new method that harnesses enzymes, offering initial clues as to how these technical obstacles may be overcome. Their findings are reported in Nature, the 20 October 2022.

Nature has invented the best solution for storing a massive amount of data: DNA. This understanding has inspired the use of DNA for the storage of digital data, converting binary (0 or 1) numbers into one of the four different DNA ‘letters’ (A, T, C, or G). 

But how does one find a specific datum in the library of information stored as DNA? And how can calculations with DNA-encoded data be performed directly, without first converting them back into electronic data? These are the questions that teams from the LIMMS (CNRS/University of Tokyo) and Gulliver (CNRS/ESPCI Paris-PSL) research laboratories have sought to answer. They are testing a new approach using enzymes and applying the solutions of artificial neurons and neural networks for direct operations on DNA data.

Specifically, the researchers have made use of the reactions of three enzymes to design chemical ‘neurons’ that reproduce the network architecture and ability for complex calculations exhibited by true neurons. Their chemical neurons can execute calculations with data on DNA strands and express the results as fluorescent signals.

The LIMMS and Gulliver teams have also innovated by assembling two layers of the artificial neurons in order to refine calculations. Precision is further enhanced through microfluidic miniaturization of reactions, allowing tens of thousands to take place.

The fruit of a decade of cooperation between French biochemists and Japanese microfluidics engineers, these breakthroughs may eventually permit better screening for certain diseases as well as the manipulation of gigantic DNA-encoded databases. 

When kept away from water, air, and light, DNA can be preserved for hundreds of thousands of years, without any energy input. And stored in a capsule a few centimetres in diameter, it can hold up to 500TB of digital data. By 2025, the total volume of digital data generated by humans is expected to reach 175ZB. Since current storage media are relatively bulky, fragile, and energy-intensive, DNA may provide a viable alternative – able to contain all existing data within the space of a shoebox. Facilitating DNA storage will be the PEPR MoleculArxiv goal, a prority research program provided last May by the CNRS.

Article: Nonlinear decision-making with enzymatic neural networks

Nature has published an article written by S. Okumura, LIMMS, CNRS-Institute of Industrial Science, University of Tokyo, Tokyo, Japan, G. Gines, Laboratoire Gulliver, PSL Research University, Paris, France, N. Lobato-Dauzier, A. Baccouche, R. Deteix, T. Fujii, LIMMS, CNRS-Institute of Industrial Science, University of Tokyo, Y. Rondelez, Laboratoire Gulliver, PSL Research University, Paris, France, and A. J. Genotn LIMMS, CNRS-Institute of Industrial Science, University of Tokyo, Tokyo, Japan.

Abstract: “Artificial neural networks have revolutionized electronic computing. Similarly, molecular networks with neuromorphic architectures may enable molecular decision-making on a level comparable to gene regulatory networks. Non-enzymatic networks could in principle support neuromorphic architectures, and seminal proofs-of-principle have been reported. However, leakages (that is, the unwanted release of species), as well as issues with sensitivity, speed, preparation and the lack of strong nonlinear responses, make the composition of layers delicate, and molecular classifications equivalent to a multilayer neural network remain elusive (for example, the partitioning of a concentration space into regions that cannot be linearly separated). Here we introduce DNA-encoded enzymatic neurons with tuneable weights and biases, and which are assembled in multilayer architectures to classify nonlinearly separable regions. We first leverage the sharp decision margin of a neuron to compute various majority functions on 10 bits. We then compose neurons into a two-layer network and synthetize a parametric family of rectangular functions on a microRNA input. Finally, we connect neural and logical computations into a hybrid circuit that recursively partitions a concentration plane according to a decision tree in cell-sized droplets. This computational power and extreme miniaturization open avenues to query and manage molecular systems with complex contents, such as liquid biopsies or DNA databases.“