Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Neural networks consisting of DNA

Michael te Vrugt

TL;DR

The chapter investigates neural-network-like computation in DNA, arguing that soft matter substrates can host intelligent processing. It centers on a DNA-based winner-take-all architecture and a DNA reservoir computing proposal, detailing how DNA gates, toehold-mediated strand displacement, seesaw gates, annihilators, and catalytic cycles realize input encoding, matrix multiplication, summation, and selection. It demonstrates handwritten digit recognition using a DNA neural network and discusses the potential and limitations of DNA-based AI, including slower dynamics but strong parallelism and tunability. Overall, the work provides a foundational, pedagogical account of intelligent matter built from DNA with potential biomedical and educational applications.

Abstract

Neural networks based on soft and biological matter constitute an interesting potential alternative to traditional implementations based on electric circuits. DNA is a particularly promising system in this context due its natural ability to store information. In recent years, researchers have started to construct neural networks that are based on DNA. In this chapter, I provide a very basic introduction to the concept of DNA neural networks, aiming at an audience that is not familiar with biochemistry.

Neural networks consisting of DNA

TL;DR

The chapter investigates neural-network-like computation in DNA, arguing that soft matter substrates can host intelligent processing. It centers on a DNA-based winner-take-all architecture and a DNA reservoir computing proposal, detailing how DNA gates, toehold-mediated strand displacement, seesaw gates, annihilators, and catalytic cycles realize input encoding, matrix multiplication, summation, and selection. It demonstrates handwritten digit recognition using a DNA neural network and discusses the potential and limitations of DNA-based AI, including slower dynamics but strong parallelism and tunability. Overall, the work provides a foundational, pedagogical account of intelligent matter built from DNA with potential biomedical and educational applications.

Abstract

Neural networks based on soft and biological matter constitute an interesting potential alternative to traditional implementations based on electric circuits. DNA is a particularly promising system in this context due its natural ability to store information. In recent years, researchers have started to construct neural networks that are based on DNA. In this chapter, I provide a very basic introduction to the concept of DNA neural networks, aiming at an audience that is not familiar with biochemistry.
Paper Structure (10 sections, 4 equations, 2 figures)

This paper contains 10 sections, 4 equations, 2 figures.

Table of Contents

  1. Introduction
  2. Biochemistry for beginners
  3. DNA Computing
  4. The winner takes it all
  5. DNA gates
  6. A DNA neural network
  7. DNA reservoir computing
  8. Advantages and disadvantages of this approach
  9. Summary
  10. Acknowledgements

Figures (2)

  • Figure 1: Visualization of toehold exchange. (A similar illustration can be found in Ref. QianW2011b.)
  • Figure 2: Feeding a hand-written letter (L) as an input signal in the form of DNA into the system. The nine pixels are represented by nine distinct DNA molecules. An L is encoded by the presence of A1, B1, C1, C2, and C3. (A similar illustration can be found in Fig. 1b of Ref. CherryQ2018 and in Fig. 2b of Ref. teVrugt2024.)