Future DNA computing device and accompanied tool stack: Towards high-throughput computation

https://doi.org/10.1016/j.future.2020.10.038Get rights and content

Highlights

  • Gives a comprehensive review of DNA computing technology.

  • Propose an architecture of DNA computing device for high-throughput computation.

  • Propose the solutions for various key component design.

  • Propose a prototype of DNA compiler for the programming on DNA computing devices.

Abstract

DNA Computing is still at its infant stage since its emergence. Multiple aspects of DNA computing have been studied but most of the research results have not been applied to the reality. It has been proved to exhibit high data storage density and support efficient random data access. It also shows the potential to provide alternative facilities for general computing. Especially, its natural double-helix structure enables it to be the best fit for high-throughput parallel computing. However, the underlying rationale of DNA computing is different from the existing electronic computing devices. Therefore, the popularity of DNA computing device is limited by its accessibility. We propose our designs for high-throughput DNA computing including DNA computing circuits, system architecture, and its compiler. We also demonstrate its feasibility using a simple example through simulation experiments. The objective of this framework is to bridge the gap between the existing computer developer community and the DNA computing biotechnology.

Introduction

DNA computing is a thrilling technology in computing. Instead of silicon-based electronic circuits, DNA computing utilizes molecular reaction techniques to compute on DNA molecules. DNA computing can be dated back to 1990s when Adleman proposed the computational model to solve the 7-node Hamiltonian path problem using DNA molecules [1]. This is the first attempt to solve mathematical problems with molecular reaction techniques. Afterwards, his group tried to solve the 20-variable 3-SAT problem with DNA computing [2]. Even though the idea is novel, the relevant application is still limited due to the customized design of the model merely for one specific problem [3], [4]. A different model proposed by Winfree aims for solving the counting problems using DNA molecules to mimic the Wang tiles [5]. Instead of exhaustive enumeration and filtering, this model exploited the structural advantages of DNA computing by converting the problem into a binary computation which is then mapped to DNA molecular interactions. However, this model is also limited by its customized application scope.

The emergence of DNA automaton promoted the development of general DNA computing techniques [6]. By defining a set of data storage steps that transform digital data into DNA sequence patterns and the accompanying operations, the DNA molecules can be used to implement an automata for general computing [7], [8]. Since then, DNA computing started to be demonstrated for its unprecedented potential in general computation. It could be used to construct the substitute parts of the existing electronic circuits for general computing.

General DNA computation became the main trend among DNA computing researchers in the early 21st century. Specifically, DNA walker is firstly proposed by Yin et al. for information transduction along DNA tracks [9]. Similar to the electric charge walker [9], the tracks of DNA walkers can be designed to implement the integrated circuits for general or specific purposes [10]. However, the DNA walker is explicitly implemented to mimic the behavior of electric circuits without making full use of the natural superiority of DNA computing features such as its high data density and parallel reaction capability.

Few years later, the DNA strand displacement (DSD) technology led the development of DNA computing to a new stage [11]. Built upon DNA hybridization, the DSD technology transforms data information through designed DSD reactions. A specific operation (i.e. logical AND, logical OR, numeric ADD, etc.) can be implemented by combining a set of DNA reactions. This technology fully utilizes the superiority of DNA computing characteristics (e.g. the high parallelism as well as storage density). It also helps building reusable components for DNA computing; it is equivalent to the logical modules of large-scale integrated circuits. Since then, researchers have attempted to construct neural network hardware using DSD components [12]. This can be a promising trend for the application of DNA computing in light of the weakness of DNA computing such as the long warming-up time and the compatibility with current computing techniques. Furthermore, deep learning tends to become one of the most popular techniques in the near future. However, it requires a large amount of computing resources and energy. Therefore, DNA computing seems to be an ideal alternative technique that could overcome those shortcomings [13].

Recently, several DNA computing systems are proposed in the literature, making this area appeals more attentions than ever before. Switching circuits (SC) is firstly proposed by Shannon [14] and now applied to digital computing system building [15]. SC is characterized for its fast speed and high bandwidth. However, compared with DSD, it has longer transmission delay and lower error tolerance. Another simplified design is proposed to achieve a faster speed in calculations [16]. But it is constructed only by AND and OR gates which lack diversity in constructing systems. The technique strand displacement synthesis is also used in developing computing system [17]. It is more flexible in constructing the systems, but it requires more space to store the synthesis sequences; it would not be a problem in DSD-based system.

Currently, the traditional silicon-based computing works well for most of the sequential tasks. However, the increasing demand in parallel computing on both computer center or consumer devices results in different problems such as energy efficiency [18], [19]. DNA computing is suitable for parallel computing, thanks to its high throughput computing and high-density data storage efficiency. On the other hand, its requirement for running environment and the compatibility with the existing computing tool stacks limits its development and wide adaptation. As a result, DNA computing cannot become the full replacement of silicon-based computing but an auxiliary system that is responsible for large-scale parallel computing on high-volume data in a stochastic manner.

Although the research area of DNA computing has beensteadily developing in recent years, its potential impacts on global computing platforms remain obscure. Therefore, we introduced its detailed techniques in the following sections and proposed a prototype system which adopts DNA computing for building a neural network model. The proposed system design could shed lights on understanding how this technology can contribute to general-purpose computation with examples. The rest of this paper is arranged as follows: the second section gives a comprehensive review of the concrete techniques and rationale in DNA computing; the third section demonstrates the architecture design of the proposed DNA computing device, its software stack, and the simulation experiments; the fourth section summarizes the above content and explores the potential future work.

Section snippets

Background

The DNA computing technology relies on the DNA reaction mechanism such as DSD [11], DNA walker [10], programmable self-assembly [20], and others [21], [22]. Those mechanism reveals the underlying components of digital logic circuits in DNA computing and the hierarchy they scale up to form the integrated circuits [23]. In particular, the lasted developed techniques are mostly based on the DSD mechanism for its robustness in the stochastic nature of molecular reactions [24].

System design

The majority of the existing studies about DNA computing focus on a specific component design (e.g. adder, counter, etc.) or for a specific problem (e.g. Hamiltonian path). However, the ultimate objective of those techniques is to build a complete system that could execute general-purpose computing tasks. The system currently does not aim at replacing all the existing silicon-based computing devices since it would take time for researchers and developers to get used to DNA computing. There is

Discussion

The simulation experiments demonstrate the potential of DNA computing to replace silicon-based computing for intensive computational tasks. The high-throughput parallel computing ability and the high-density storage capacity of DNA computing devices can enable large-scale parallel computing on big data. As illustrated in our study, the DNA strand displacement and the digital circuits with DSD can be the fundamental building blocks while the neural network architecture plots the overall

Conclusion

In this work, we reviewed the development and the typical designs of DNA computing systems. We also gave an overview on the existing biotechnologies for DNA computing. We elaborated the details of those techniques and reveal the principle of computing driven by those biotechnologies. Afterwards, we proposed our prototype DNA computing system including the architecture, coding scheme, and compiler. Those crucial components lay out the foundation of integrated DNA computing system. The prototype

CRediT authorship contribution statement

Shankai Yan: Completed most of the research work in this paper including experiment design and conducting, Draft writing. Ka-Chun Wong: Review proofreading, Draft editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

We thank the two anonymous reviewers for their careful and thoughtful comments which have improved the manuscript reader-friendliness in a significant manner.

Shankai Yan received his Ph.D. degree in Department of Computer Science at City University of Hong Kong (4th in QS World University Ranking aged under 50 and 5th in Times University Ranking aged under 50) under the supervision of Dr. WONG Ka-Chun. After that, he was awarded fellowship of NIH and work as a Postdoctoral Fellow in Zhiyong Lu’s BioNLP Lab at NCBI/NLM/NIH. He has papers with self-selected topics published on Bioinformatics and JBHI, demonstrating solid research skills during his

References (36)

  • AbualrubT. et al.

    Construction of cyclic codes over gf(4) for dna computing

    J. Franklin Inst. B

    (2006)
  • AdlemanL.

    Molecular computation of solutions to combinatorial problems

    Science

    (1994)
  • BraichR.S. et al.

    Solution of a 20-variable 3-SAT problem on a DNA computer

    Science

    (2002)
  • IgnatovaZ. et al.

    DNA Computing Models

    (2008)
  • XuJ. et al.

    A review on DNA computing models

    J. Comput. Theor. Nanosci.

    (2007)
  • WinfreeE. et al.

    Design and self-assembly of two-dimensional dna crystals

    Nature

    (1998)
  • BenensonY. et al.

    Programmable and autonomous computing machine made of biomolecules

    Nature

    (2001)
  • HjelmfeltY. et al.

    Chemical implementation of finite-state machines

    Proc. Natl. Acad. Sci.

    (1992)
  • Von NeumannJ. et al.

    Theory of self-reproducing automata

    IEEE Trans. Neural Netw.

    (1966)
  • YinP. et al.

    A unidirectional DNA walker that moves autonomously along a track

    Angew. Chem.

    (2004)
  • DannenbergF. et al.

    DNA Walker circuits: Computational potential, design, and verification

  • ZhangD.Y. et al.

    Control of DNA strand displacement kinetics using toehold exchange

    J. Am. Chem. Soc.

    (2009)
  • QianL. et al.

    Scaling up digital circuit computation with DNA strand displacement cascades

    Science

    (2011)
  • HjelmfeltA. et al.

    Chemical implementation of neural networks and turing machines

    Proc. Natl. Acad. Sci. USA

    (1991)
  • ShannonC.E.

    A symbolic analysis of relay and switching circuits

    Electr. Eng.

    (2013)
  • WangF. et al.

    Implementing digital computing with DNA-based switching circuits

    Nature Commun.

    (2020)
  • SongT. et al.

    Fast and compact DNA logic circuits based on single-stranded gates using strand-displacing polymerase

    Nature Nanotechnol.

    (2019)
  • SuH. et al.

    High-efficiency and integrable DNA arithmetic and logic system based on strand displacement synthesis

    Nature Commun.

    (2019)
  • Cited by (7)

    View all citing articles on Scopus

    Shankai Yan received his Ph.D. degree in Department of Computer Science at City University of Hong Kong (4th in QS World University Ranking aged under 50 and 5th in Times University Ranking aged under 50) under the supervision of Dr. WONG Ka-Chun. After that, he was awarded fellowship of NIH and work as a Postdoctoral Fellow in Zhiyong Lu’s BioNLP Lab at NCBI/NLM/NIH. He has papers with self-selected topics published on Bioinformatics and JBHI, demonstrating solid research skills during his Ph.D. study. His research is mainly about applied machine learning in biomedical fields. He is skilled in designing and coding machine/deep learning models with high efficiency.

    View full text