Scaling down DNA circuits with competitive neural networks

Harnessing DNA computing and nanopore decoding for practical applications: from informatics to microRNA-targeting diagnostics

DNA computing represents a subfield of molecular computing with the potential to become a significant area of next-generation computation due to the high programmability inherent in the sequence-dependent molecular behaviour of DNA. Recent studies in DNA computing have extended from mathematical informatics to biomedical applications, with a particular focus on diagnostics that exploit the biocompatibility of DNA molecules. The output of DNA computing devices is encoded in nucleic acid molecules, which must then be decoded into human-recognizable signals for practical applications. Nanopore technology, which utilizes an electrical and label-free decoding approach, provides a unique platform to bridge DNA and electronic computing for practical use. In this tutorial review, we summarise the fundamental knowledge, technologies, and methodologies of DNA computing (logic gates, circuits, neural networks, and non-DNA input circuity). We then focus on nanopore-based decoding, and highlight recent advances in medical diagnostics targeting microRNAs as biomarkers. Finally, we conclude with the potential and challenges for the practical implementation of these techniques. We hope that this tutorial will provide a comprehensive insight and enable the general reader to grasp the fundamental principles and diverse applications of DNA computing and nanopore decoding, and will inspire a wide range of scientists to explore and push the boundaries of these technologies.

Refined design of a DNA logic gate for implementing a DNA-based three-level circuit

DNA computing circuits are favored by researchers because of their high density, high parallelism, and biocompatibility. However, compared with electronic circuits, current DNA circuits have significant errors in understanding the OFF state and logic "0". Nowadays, DNA circuits only have two input states: logic "0" and logic "1", where logic "0" also means the OFF state. Corresponding to an electronic circuit, it is more like an on-off switch than a logic circuit. To correct this conceptual confusion, we propose a three-level circuit. The circuit divides the input signal into three cases: "none", logic "0" and logic "1". In subsequent experiments, 34 input combinations of the primary AND gate, OR gate as well as secondary AND-OR and OR-AND cascade circuits were successfully implemented to perform the operation, which distinguished the OFF state and logic "0" correctly. Based on this, we successfully implemented a more complex voting operation with only 12 strands. We believe that our redefinition of the OFF state and logic "0" will promote tremendous developments in DNA computing circuits.

A synthetic protein-level neural network in mammalian cells

Artificial neural networks provide a powerful paradigm for nonbiological information processing. To understand whether similar principles could enable computation within living cells, we combined de novo-designed protein heterodimers and engineered viral proteases to implement a synthetic protein circuit that performs winner-take-all neural network classification. This "perceptein" circuit combines weighted input summation through reversible binding interactions with self-activation and mutual inhibition through irreversible proteolytic cleavage. These interactions collectively generate a large repertoire of distinct protein species stemming from up to eight coexpressed starting protein species. The complete system achieves multi-output signal classification with tunable decision boundaries in mammalian cells and can be used to conditionally control cell death. These results demonstrate how engineered protein-based networks can enable programmable signal classification in living cells.

A novel activation function based on DNA enzyme-free hybridization reaction and its implementation on nonlinear molecular learning systems

With the advent of the post-Moore's Law era, the development of traditional silicon-based computers has reached its limit, and there is an urgent need to develop new computing technologies to meet the needs of science, technology, and daily life. Due to its super-strong parallel computing capability and outstanding data storage capacity, DNA computing has become an important branch and hot research topic of new computer technology. DNA enzyme-free hybridization reaction technology is widely used in DNA computing, showing excellent performance in computing power and information processing. Studies have shown that DNA molecules not only have the computing function of electronic devices, but also exhibit certain human brain-like functions. In the field of artificial intelligence, activation functions play an important role as they enable artificial intelligence systems to fit and predict non-linear and complex variable relationships. Due to the difficulty of implementing activation functions in DNA computing, DNA circuits cannot easily achieve all the functions of artificial intelligence. DNA circuits need to rely on electronic computers to complete the training and learning process. Based on the parallel computing characteristics of DNA computing and the kinetic features of DNA molecule displacement reactions, this paper proposes a new activation function. This activation function can not only be easily implemented by DNA enzyme-free hybridization reaction reactions, but also has good nesting properties in DNA circuits, and can be cascaded with other DNA reactions to form a complete DNA circuit. This paper not only provides the mathematical analysis of the proposed activation function, but also provides a detailed analysis of its kinetic features. The activation function is then nested into a nonlinear neural network for DNA computing. This system is capable of fitting and predicting a certain nonlinear function.

Synthetic biological neural networks: From current implementations to future perspectives

Artificial neural networks, inspired by the biological networks of the human brain, have become game-changing computing models in modern computer science. Inspired by their wide scope of applications, synthetic biology strives to create their biological counterparts, which we denote synthetic biological neural networks (SYNBIONNs). Their use in the fields of medicine, biosensors, biotechnology, and many more shows great potential and presents exciting possibilities. So far, many different synthetic biological networks have been successfully constructed, however, SYNBIONN implementations have been sparse. The latter are mostly based on neural networks pretrained in silico and being heavily dependent on extensive human input. In this paper, we review current implementations and models of SYNBIONNs. We briefly present the biological platforms that show potential for designing and constructing perceptrons and/or multilayer SYNBIONNs. We explore their future possibilities along with the challenges that must be overcome to successfully implement a scalable in vivo biological neural network capable of online learning.

A molecular assessment of the practical potential of DNA-based computation

The immense information density of DNA and its potential for massively parallelized computations, paired with rapidly expanding data production and storage needs, have fueled a renewed interest in DNA-based computation. Since the construction of the first DNA computing systems in the 1990s, the field has grown to encompass a diverse array of configurations. Simple enzymatic and hybridization reactions to solve small combinatorial problems transitioned to synthetic circuits mimicking gene regulatory networks and DNA-only logic circuits based on strand displacement cascades. These have formed the foundations of neural networks and diagnostic tools that aim to bring molecular computation to practical scales and applications. Considering these great leaps in system complexity as well as in the tools and technologies enabling them, a reassessment of the potential of such DNA computing systems is warranted.

A survey on molecular-scale learning systems with relevance to DNA computing

DNA computing has emerged as a promising alternative to achieve programmable behaviors in chemistry by repurposing the nucleic acid molecules into chemical hardware upon which synthetic chemical programs can be executed. These chemical programs are capable of simulating diverse behaviors, including boolean logic computation, oscillations, and nanorobotics. Chemical environments such as the cell are marked by uncertainty and are prone to random fluctuations. For this reason, potential DNA-based molecular devices that aim to be deployed into such environments should be capable of adapting to the stochasticity inherent in them. In keeping with this goal, a new subfield has emerged within DNA computing, focusing on developing approaches that embed learning and inference into chemical reaction systems. If realized in biochemical contexts, such molecular machines can engender novel applications in fields such as biotechnology, synthetic biology, and medicine. Therefore, it would be beneficial to review how different ideas were conceived, how the progress has been so far, and what the emerging ideas are in this nascent field of 'molecular-scale learning'.

Molecular Computation for Molecular Classification

DNA as an informational polymer has, for the past 30 years, progressively become an essential molecule to rationally build chemical reaction networks endowed with powerful signal-processing capabilities. Whether influenced by the silicon world or inspired by natural computation, molecular programming has gained attention for diagnosis applications. Of particular interest for this review, molecular classifiers have shown promising results for disease pattern recognition and sample classification. Because both input integration and computation are performed in a single tube, at the molecular level, this low-cost approach may come as a complementary tool to molecular profiling strategies, where all biomarkers are quantified independently using high-tech instrumentation. After introducing the elementary components of molecular classifiers, some of their experimental implementations are discussed either using digital Boolean logic or analog neural network architectures.

Positive feedback drives a secondary nonlinear product burst during a biphasic DNA amplification reaction

Isothermal DNA amplification reactions are used in a broad variety of applications, from diagnostic assays to DNA circuits, with greater speed and less complexity than established PCR technologies. We recently reported a unique, high gain, biphasic isothermal DNA amplification reaction, called the Ultrasensitive DNA Amplification Reaction (UDAR). Here we present a detailed analysis of the UDAR reaction pathways that initiates with a first phase followed by a nonlinear product burst, which is caused by an autocatalytic secondary reaction. The experimental reaction output was reproduced using an ordinary differential equation model based on detailed reaction mechanisms. This model provides insight on the relative importance of each reaction mechanism during both phases, which could aid in the design of product output during DNA amplification reactions.

Nonlinear decision-making with enzymatic neural networks

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^1,2. Non-enzymatic networks could in principle support neuromorphic architectures, and seminal proofs-of-principle have been reported^3,4. 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.

Massively Parallel DNA Computing Based on Domino DNA Strand Displacement Logic Gates

DNA computing has gained considerable attention due to the characteristics of high-density information storage and high parallel computing for solving computational problems. Building addressable logic gates with biomolecules is the basis for establishing biological computers. In the current calculation model, the multiinput AND operation often needs to be realized through a multilevel cascade between logic gates. Through experiments, it was found that the multilevel cascade causes signal leakage and affects the stability of the system. Using DNA strand displacement technology, we constructed a domino-like multiinput AND gate computing system instead of a cascade of operations, realizing multiinput AND computing on one logic gate and abandoning the traditional multilevel cascade of operations. Fluorescence experiments demonstrated that our methods significantly reduce system construction costs and improve the stability and robustness of the system. Finally, we proved stability and robustness of the domino AND gate by simulating the tic-tac-toe process with a massively parallel computing strategy.

Programming Molecular Systems To Emulate a Learning Spiking Neuron

Hebbian theory seeks to explain how the neurons in the brain adapt to stimuli to enable learning. An interesting feature of Hebbian learning is that it is an unsupervised method and, as such, does not require feedback, making it suitable in contexts where systems have to learn autonomously. This paper explores how molecular systems can be designed to show such protointelligent behaviors and proposes the first chemical reaction network (CRN) that can exhibit autonomous Hebbian learning across arbitrarily many input channels. The system emulates a spiking neuron, and we demonstrate that it can learn statistical biases of incoming inputs. The basic CRN is a minimal, thermodynamically plausible set of microreversible chemical equations that can be analyzed with respect to their energy requirements. However, to explore how such chemical systems might be engineered de novo, we also propose an extended version based on enzyme-driven compartmentalized reactions. Finally, we show how a purely DNA system, built upon the paradigm of DNA strand displacement, can realize neuronal dynamics. Our analysis provides a compelling blueprint for exploring autonomous learning in biological settings, bringing us closer to realizing real synthetic biological intelligence.

DNA Input Classification by a Riboregulator-Based Cell-Free Perceptron

The ability to recognize molecular patterns is essential for the continued survival of biological organisms, allowing them to sense and respond to their immediate environment. The design of synthetic gene-based classifiers has been explored previously; however, prior strategies have focused primarily on DNA strand-displacement reactions. Here, we present a synthetic in vitro transcription and translation (TXTL)-based perceptron consisting of a weighted sum operation (WSO) coupled to a downstream thresholding function. We demonstrate the application of toehold switch riboregulators to construct a TXTL-based WSO circuit that converts DNA inputs into a GFP output, the concentration of which correlates to the input pattern and the corresponding weights. We exploit the modular nature of the WSO circuit by changing the output protein to the Escherichia coli σ28-factor, facilitating the coupling of the WSO output to a downstream reporter network. The subsequent introduction of a σ28 inhibitor enabled thresholding of the WSO output such that the expression of the downstream reporter protein occurs only when the produced σ28 exceeds this threshold. In this manner, we demonstrate a genetically implemented perceptron capable of binary classification, i.e., the expression of a single output protein only when the desired minimum number of inputs is exceeded.

A Loser-Take-All DNA Circuit

DNA-based neural networks are a type of DNA circuit capable of molecular pattern recognition tasks. Winner-take-all DNA networks have been developed to scale up the complexity of molecular pattern recognition with a simple molecular implementation. This simplicity was achieved by replacing negative weights in individual neurons with lateral inhibition and competition across neurons, eliminating the need for dual-rail representation. Here we introduce a new type of DNA circuit that is called loser-take-all: an output signal is ON if and only if the corresponding input has the smallest analog value among all inputs. We develop a DNA strand-displacement implementation of loser-take-all circuits that is cascadable without dual-rail representation, maintaining the simplicity desired for scalability. We characterize the impact of effective signal concentrations and reaction rates on the circuit performance, and derive solutions for compensating undesired signal loss and rate differences. Using these approaches, we successfully demonstrate a three-input loser-take-all circuit with nine unique input combinations. Complementary to winner-take-all, loser-take-all DNA circuits could be used for recognition of molecular patterns based on their least similarities to a set of memories, allowing classification decisions for patterns that are extremely noisy. Moreover, the design principle of loser-take-all could be more generally applied in other DNA circuit implementations including k-winner-take-all.

Biocomputing Based on DNA Strand Displacement Reactions

The high sequence specificity and precise base complementary pairing principle of DNA provides a rich orthogonal molecular library for molecular programming, making it one of the most promising materials for developing bio-compatible intelligence. In recent years, DNA has been extensively studied and applied in the field of biological computing. Among them, the toehold-mediated strand displacement reaction (SDR) with properties including enzyme free, flexible design and precise control, have been extensively used to construct biological computing circuits. This review provides a systemic overview of SDR design principles and the applications. Strategies for designing DNA-only, enzymes-assisted, other molecules-involved and external stimuli-controlled SDRs are described. The recently realized computing functions and the application of DNA computing in other fields are introduced. Finally, the advantages and challenges of SDR-based computing are discussed.

Multiplex Digital MicroRNA Detection Using Cross-Inhibitory DNA Circuits

Ubiquitous post-transcriptional regulators in eukaryotes, microRNAs are currently emerging as promising biomarkers of physiological and pathological processes. Multiplex and digital detection of microRNAs represents a major challenge toward the use of microRNA signatures in clinical settings. The classical reverse transcription polymerase chain reaction quantification approach has important limitations because of the need for thermocycling and a reverse transcription step. Simpler, isothermal alternatives have been proposed, yet none could be adapted in both a digital and multiplex format. This is either because of a lack of sensitivity that forbids single molecule detection or molecular cross-talk reactions that are responsible for nonspecific amplification. Building on an ultrasensitive isothermal amplification mechanism, we present a strategy to suppress cross-talk reactions, allowing for robust isothermal and multiplex detection of microRNA targets. Our approach relies on target-specific DNA circuits interconnected with DNA-encoded inhibitors that repress nonspecific signal amplification. We demonstrate the one-step, isothermal, digital, and simultaneous quantification of various pairs of important microRNA targets.

Reinforcement learning in synthetic gene circuits

Synthetic gene circuits allow programming in DNA the expression of a phenotype at a given environmental condition. The recent integration of memory systems with gene circuits opens the door to their adaptation to new conditions and their re-programming. This lays the foundation to emulate neuromorphic behaviour and solve complex problems similarly to artificial neural networks. Cellular products such as DNA or proteins can be used to store memory in both digital and analog formats, allowing cells to be turned into living computing devices able to record information regarding their previous states. In particular, synthetic gene circuits with memory can be engineered into living systems to allow their adaptation through reinforcement learning. The development of gene circuits able to adapt through reinforcement learning moves Sciences towards the ambitious goal: the bottom-up creation of a fully fledged living artificial intelligence.

Domain-Specific Programming Languages for Computational Nucleic Acid Systems

The construction of models of system behavior is of great importance throughout science and engineering. In bioengineering and bionanotechnology, these often take the form of dynamic models that specify the evolution of different species over time. To ensure that scientific observations and conclusions are consistent and that systems can be reliably engineered on the basis of model predictions, it is important that models of biomolecular systems can be constructed in a reliable, principled, and efficient manner. This review focuses on efforts to address this need by using domain-specific programming languages as the basis for custom design tools for researchers working on computational nucleic acid devices, where a domain-specific language is simply a programming language tailored to a particular application domain. The underlying thesis of our review is that there is a continuum of practical implementation strategies for computational nucleic acid systems, which can all benefit from appropriate domain-specific languages and software design tools. We emphasize the need for specialized yet flexible tools that can be realized using domain-specific languages that compile to more general-purpose representations.

On the importance of reaction networks for synthetic living systems

The goal of creating a synthetic cell necessitates the development of reaction networks which will underlie all of its behaviours. Recent developments in in vitro systems, based upon both DNA and enzymes, have created networks capable of a range of behaviours e.g. information processing, adaptation and diffusive signalling. These networks are based upon reaction motifs that when combined together produce more complex behaviour. We highlight why it is inevitable that networks, based on enzymes or enzyme-like catalysts, will be required for the construction of a synthetic cell. We outline several of the challenges, including (a) timing, (b) regulation and (c) energy distribution, that must be overcome in order to transition from the simple networks we have today to much more complex networks capable of a variety of behaviours and which could find application one day within a synthetic cell.

Automated sequence-level analysis of kinetics and thermodynamics for domain-level DNA strand-displacement systems

As an engineering material, DNA is well suited for the construction of biochemical circuits and systems, because it is simple enough that its interactions can be rationally designed using Watson-Crick base pairing rules, yet the design space is remarkably rich. When designing DNA systems, this simplicity permits using functional sections of each strand, called domains, without considering particular nucleotide sequences. However, the actual sequences used may have interactions not predicted at the domain-level abstraction, and new rigorous analysis techniques are needed to determine the extent to which the chosen sequences conform to the system's domain-level description. We have developed a computational method for verifying sequence-level systems by identifying discrepancies between the domain-level and sequence-level behaviour. This method takes a DNA system, as specified using the domain-level tool Peppercorn, and analyses data from the stochastic sequence-level simulator Multistrand and sequence-level thermodynamic analysis tool NUPACK to estimate important aspects of the system, such as reaction rate constants and secondary structure formation. These techniques, implemented as the Python package KinDA, will allow researchers to predict the kinetic and thermodynamic behaviour of domain-level systems after sequence assignment, as well as to detect violations of the intended behaviour.

Scaling up molecular pattern recognition with DNA-based winner-take-all neural networks

From bacteria following simple chemical gradients¹ to the brain distinguishing complex odour information², the ability to recognize molecular patterns is essential for biological organisms. This type of information-processing function has been implemented using DNA-based neural networks³, but has been limited to the recognition of a set of no more than four patterns, each composed of four distinct DNA molecules. Winner-take-all computation⁴ has been suggested^5,6 as a potential strategy for enhancing the capability of DNA-based neural networks. Compared to the linear-threshold circuits⁷ and Hopfield networks⁸ used previously³, winner-take-all circuits are computationally more powerful⁴, allow simpler molecular implementation and are not constrained by the number of patterns and their complexity, so both a large number of simple patterns and a small number of complex patterns can be recognized. Here we report a systematic implementation of winner-take-all neural networks based on DNA-strand-displacement^9,10 reactions. We use a previously developed seesaw DNA gate motif^3,11,12, extended to include a simple and robust component that facilitates the cooperative hybridization¹³ that is involved in the process of selecting a 'winner'. We show that with this extended seesaw motif DNA-based neural networks can classify patterns into up to nine categories. Each of these patterns consists of 20 distinct DNA molecules chosen from the set of 100 that represents the 100 bits in 10 × 10 patterns, with the 20 DNA molecules selected tracing one of the handwritten digits '1' to '9'. The network successfully classified test patterns with up to 30 of the 100 bits flipped relative to the digit patterns 'remembered' during training, suggesting that molecular circuits can robustly accomplish the sophisticated task of classifying highly complex and noisy information on the basis of similarity to a memory.

A molecular multi-gene classifier for disease diagnostics

Despite its early promise as a diagnostic and prognostic tool, gene expression profiling remains cost-prohibitive and challenging to implement in a clinical setting. Here, we introduce a molecular computation strategy for analysing the information contained in complex gene expression signatures without the need for costly instrumentation. Our workflow begins by training a computational classifier on labelled gene expression data. This in silico classifier is then realized at the molecular level to enable expression analysis and classification of previously uncharacterized samples. Classification occurs through a series of molecular interactions between RNA inputs and engineered DNA probes designed to differentially weigh each input according to its importance. We validate our technology with two applications: a classifier for early cancer diagnostics and a classifier for differentiating viral and bacterial respiratory infections based on host gene expression. Together, our results demonstrate a general and modular framework for low-cost gene expression analysis.

Solving probability reasoning based on DNA strand displacement and probability modules

In computation biology, DNA strand displacement technology is used to simulate the computation process and has shown strong computing ability. Most researchers use it to solve logic problems, but it is only rarely used in probabilistic reasoning. To process probabilistic reasoning, a conditional probability derivation model and total probability model based on DNA strand displacement were established in this paper. The models were assessed through the game "read your mind." It has been shown to enable the application of probabilistic reasoning in genetic diagnosis.

Microscopic agents programmed by DNA circuits

Information stored in synthetic nucleic acids sequences can be used in vitro to create complex reaction networks with precisely programmed chemical dynamics. Here, we scale up this approach to program networks of microscopic particles (agents) dispersed in an enzymatic solution. Agents may possess multiple stable states, thus maintaining a memory and communicate by emitting various orthogonal chemical signals, while also sensing the behaviour of neighbouring agents. Using this approach, we can produce collective behaviours involving thousands of agents, for example retrieving information over long distances or creating spatial patterns. Our systems recapitulate some fundamental mechanisms of distributed decision making and morphogenesis among living organisms and could find applications in cases where many individual clues need to be combined to reach a decision, for example in molecular diagnostics.

Materials learning from life: concepts for active, adaptive and autonomous molecular systems

A broad overview of functional aspects in biological and synthetic out-of-equilibrium systems.

Boosting functionality of synthetic DNA circuits with tailored deactivation

Molecular programming takes advantage of synthetic nucleic acid biochemistry to assemble networks of reactions, in vitro, with the double goal of better understanding cellular regulation and providing information-processing capabilities to man-made chemical systems. The function of molecular circuits is deeply related to their topological structure, but dynamical features (rate laws) also play a critical role. Here we introduce a mechanism to tune the nonlinearities associated with individual nodes of a synthetic network. This mechanism is based on programming deactivation laws using dedicated saturable pathways. We demonstrate this approach through the conversion of a single-node homoeostatic network into a bistable and reversible switch. Furthermore, we prove its generality by adding new functions to the library of reported man-made molecular devices: a system with three addressable bits of memory, and the first DNA-encoded excitable circuit. Specific saturable deactivation pathways thus greatly enrich the functional capability of a given circuit topology.

Nonlinearity in synthetic molecular circuits is usually achieved by manipulation of network topology or of production kinetics. Here, the authors achieve bistability and other nonlinear behaviours by manipulating the individual degradation rate laws of circuit components using saturable pathways.

Supervised Learning in Adaptive DNA Strand Displacement Networks

The development of engineered biochemical circuits that exhibit adaptive behavior is a key goal of synthetic biology and molecular computing. Such circuits could be used for long-term monitoring and control of biochemical systems, for instance, to prevent disease or to enable the development of artificial life. In this article, we present a framework for developing adaptive molecular circuits using buffered DNA strand displacement networks, which extend existing DNA strand displacement circuit architectures to enable straightforward storage and modification of behavioral parameters. As a proof of concept, we use this framework to design and simulate a DNA circuit for supervised learning of a class of linear functions by stochastic gradient descent. This work highlights the potential of buffered DNA strand displacement as a powerful circuit architecture for implementing adaptive molecular systems.

Design of a biochemical circuit motif for learning linear functions

Learning and adaptive behaviour are fundamental biological processes. A key goal in the field of bioengineering is to develop biochemical circuit architectures with the ability to adapt to dynamic chemical environments. Here, we present a novel design for a biomolecular circuit capable of supervised learning of linear functions, using a model based on chemical reactions catalysed by DNAzymes. To achieve this, we propose a novel mechanism of maintaining and modifying internal state in biochemical systems, thereby advancing the state of the art in biomolecular circuit architecture. We use simulations to demonstrate that the circuit is capable of learning behaviour and assess its asymptotic learning performance, scalability and robustness to noise. Such circuits show great potential for building autonomous in vivo nanomedical devices. While such a biochemical system can tell us a great deal about the fundamentals of learning in living systems and may have broad applications in biomedicine (e.g. autonomous and adaptive drugs), it also offers some intriguing challenges and surprising behaviours from a machine learning perspective.

Bridging the Two Worlds: A Universal Interface between Enzymatic and DNA Computing Systems

Molecular computing based on enzymes or nucleic acids has attracted a great deal of attention due to the perspectives of controlling living systems in the way we control electronic computers. Enzyme-based computational systems can respond to a great variety of small molecule inputs. They have the advantage of signal amplification and highly specific recognition. DNA computing systems are most often controlled by oligonucleotide inputs/outputs and are capable of sophisticated computing as well as controlling gene expressions. Here, we developed an interface that enables communication of otherwise incompatible nucleic-acid and enzyme-computational systems. The enzymatic system processes small molecules as inputs and produces NADH as an output. The NADH output triggers electrochemical release of an oligonucleotide, which is accepted by a DNA computational system as an input. This interface is universal because the enzymatic and DNA computing systems are independent of each other in composition and complexity.

Dynamic DNA-toolbox reaction circuits: a walkthrough

In living organisms, the integration of signals from the environment and the molecular computing leading to a cellular response are orchestrated by Gene Regulatory Networks (GRN). However, the molecular complexity of in vivo genetic regulation makes it next to impossible to describe in a quantitative manner. Reproducing, in vitro, reaction networks that could mimic the architecture and behavior of in vivo networks, yet lend themselves to mathematical modeling, represents a useful strategy to understand, and even predict, the function of GRN. In this paper, we define a set of in vitro, DNA-based molecular transformations that can be linked to each other in such a way that the product of one transformation can activate or inhibit the production of one or several other DNA compounds. Therefore, these reactions can be wired in arbitrary networks. This approach provides an experimental way to reproduce the dynamic features of genetic regulation in a test tube. We introduce the rules to design the necessary DNA species, a guide to implement the chemical reactions and ways to optimize the experimental conditions. We finally show how this framework, or "DNA toolbox", can be used to generate an inversion module, though many other behaviors, including oscillators and bistable switches, can be implemented.

Computer-assisted design for scaling up systems based on DNA reaction networks

In the past few years, there have been many exciting advances in the field of molecular programming, reaching a point where implementation of non-trivial systems, such as neural networks or switchable bistable networks, is a reality. Such systems require nonlinearity, be it through signal amplification, digitalization or the generation of autonomous dynamics such as oscillations. The biochemistry of DNA systems provides such mechanisms, but assembling them in a constructive manner is still a difficult and sometimes counterintuitive process. Moreover, realistic prediction of the actual evolution of concentrations over time requires a number of side reactions, such as leaks, cross-talks or competitive interactions, to be taken into account. In this case, the design of a system targeting a given function takes much trial and error before the correct architecture can be found. To speed up this process, we have created DNA Artificial Circuits Computer-Assisted Design (DACCAD), a computer-assisted design software that supports the construction of systems for the DNA toolbox. DACCAD is ultimately aimed to design actual in vitro implementations, which is made possible by building on the experimental knowledge available on the DNA toolbox. We illustrate its effectiveness by designing various systems, from Montagne et al.'s Oligator or Padirac et al.'s bistable system to new and complex networks, including a two-bit counter or a frequency divider as well as an example of very large system encoding the game Mastermind. In the process, we highlight a variety of behaviours, such as enzymatic saturation and load effect, which would be hard to handle or even predict with a simpler model. We also show that those mechanisms, while generally seen as detrimental, can be used in a positive way, as functional part of a design. Additionally, the number of parameters included in these simulations can be large, especially in the case of complex systems. For this reason, we included the possibility to use CMA-ES, a state-of-the-art optimization algorithm that will automatically evolve parameters chosen by the user to try to match a specified behaviour. Finally, because all possible functionality cannot be captured by a single software, DACCAD includes the possibility to export a system in the synthetic biology markup language, a widely used language for describing biological reaction systems. DACCAD can be downloaded online at http://www.yannick-rondelez.com/downloads/.

In vitro regulatory models for systems biology

The reductionist approach has revolutionized biology in the past 50 years. Yet its limits are being felt as the complexity of cellular interactions is gradually revealed by high-throughput technology. In order to make sense of the deluge of "omic data", a hypothesis-driven view is needed to understand how biomolecular interactions shape cellular networks. We review recent efforts aimed at building in vitro biochemical networks that reproduce the flow of genetic regulation. We highlight how those efforts have culminated in the rational construction of biochemical oscillators and bistable memories in test tubes. We also recapitulate the lessons learned about in vivo biochemical circuits such as the importance of delays and competition, the links between topology and kinetics, as well as the intriguing resemblance between cellular reaction networks and ecosystems.