Abstract:

DNA computing and DNA data storage are emerging fields that are unlocking new possibilities in information technology and diagnostics. These approaches use DNA molecules as a computing substrate or a storage medium, offering nanoscale compactness and operation in unconventional media (including aqueous solutions, water-in-oil microemulsions and self-assembled membranized compartments) for applications beyond traditional silicon-based computing systems. To build a functional DNA computer that can process and store molecular information necessitates the continued development of strategies for computing and data storage, as well as bridging the gap between these fields. In this Review, we explore how DNA can be leveraged in the context of DNA computing with a focus on neural networks and compartmentalized DNA circuits. We also discuss emerging approaches to the storage of data in DNA and associated topics such as the writing, reading, retrieval and post-synthesis editing of DNA-encoded data. Finally, we provide insights into how DNA computing can be integrated with DNA data storage and explore the use of DNA for near-memory computing for future information technology and health analysis applications.

Fig. 1 a, Base pairing enables DNA to serve as a substrate for molecular information processing. b, DNA strand displacement as a basic tool for logic gates, neural networks and compartmentalized circuits. In a typical DNA strand displacement, the single-stranded DNA input binds to the double-stranded DNA gate at the toehold domain (red) and initiates displacement. Complete displacement results in release of the output. The output can serve as an input to a downstream reaction, which allows for the connection of multiple displacement reactions and the construction of DNA circuits for DNA computing, such as logic gates and neural networks. In addition, compartmentalization offers a strategy to reach a higher level of complexity in DNA computing. c, DNA can be used to store digital data. Digital files such as text and pictures can be represented as sequences of bits and then converted into DNA sequences. The synthesized DNA sequences function as an operatable storage medium, which can be read by sequencing techniques and decoded back to digital data.

Fig. 2  a, Convolutional neural networks implemented with synthetic DNA circuits enable the recognition of handwritten symbols. The inputs of image pixel values are transformed by the convolution kernel, and then interact sequentially with downstream summation and subtraction gates to finally generate a fluorescence signal. The intensity of the signal is normalized to an output number as a score for pattern recognition. The rectified linear unit (ReLU) is a nonlinear activation function that converts negative values to 0. b, A DNA-based winner-take-all neural network enables the detection of microRNA (miRNA) for cancer diagnosis. The miRNA inputs are multiplied, summated and subtracted sequentially, resulting in the report of a value to distinguish the disease states. c, DNA logic gates that use strand-displacement polymerase. The hybridization of inputs and single-stranded gates results in the polymerase-mediated synthesis of new strands and concomitant strand-displacement reactions, which finally results in the release of output strands from the gate complex. Individual gates can be combined by cascading strategies (that is, sequences of consecutive chemical transformations) to construct large-scale logic circuits. Examples of such circuits include fan-in circuits (which process multiple inputs into a single output), fan-out circuits (which process a single input into multiple outputs) and circuits that use square-root functions to compute four-bit input numbers. d, An enzymatic neural network with two layers that can be used to classify complex mixtures of DNA by nonlinear decision making. The concentrations of two DNA inputs (X1 and X2) are detected by enzymatic reaction networks (α and β), the results of which are integrated by a third reaction network γ, resulting in classification of the input concentrations into three areas (shown in the dashed boxes). Part a adapted from ref. 50, Springer Nature Limited. Part b adapted from ref. 52, Springer Nature Limited. Part c adapted from ref. 16, Springer Nature Limited. Part d adapted from ref. 85, Springer Nature Limited.

DOI: 10.1038/s41570-024-00576-4