Research On Computational System Based On Nucleic Acid Molecular Regulatory Network

In nature,the ability to process information is crucial to the organisms for survival and evolution,such as impressive capabilities of brain — ranging from perception,pattern recognition and decision making.Before the evolution of neuron-based brain,complex biomolecular circuits provided individual cells with the “smart” behaviors they needed to survive.The study of how molecules can ‘think’ has produced a variety of computational models and applications of artificial chemical systems,however,the development of functional circuits with more powerful information-processing capability is one of the major research goals of the field of DNA nanotechnology.Nucleic acid molecules have many desirable properties for engineering artificial molecular circuits,in which their sequences can be precisely controlled to encode different signals while crosstalk between them can be avoided,and highly specific.In this thesis,by designing and building DNA switches with responsive behavior,signal propagation and responsive regulation functions of which can be controlled independently,which can be used to independently control the chemical reaction pathways.A variety of DNA computing units have been realized by using DNA switches as building blocks.Furthermore,multi-functional DNA regulatory circuits that implement cascaded logic,analog and neural network computing have been constructed by cascading multiple units,which have demonstrated the information processing capability and potential of DNA circuits.The major research contents are as follows:(1)The design of DNA structure can be done at the domain level.By this way,the proton-responsive triplex DNA motif can be inserted between the toehold and the recognition domain to form DNA switches with different p H windows.Using one photochromic molecule that is allowed for UV-induced isomerization to influence the protons concentrations in solution,the optochemically controlled DNA switch has been realized.Based on this DNA switch,4 DNA computing units were constructed to perform logic and analog computation,and each of which was verified experimentally.(2)The modularity and scalability of the DNA computing units are important for the construction of multi-functional DNA regulatory circuits.Using UV-induced isomerization property of the photochromic molecule,the switchable multi-input cascaded logic circuits were constructed by cascading the DNA logic computing units.In addition,the triplex-based DNA switch can be used for the construction of DNA regulatory circuits that perform 2-bit and 1-bit probabilistic computation,which shows the feasibility.Moreover,using DNA analog computing units as building blocks,DNA regulatory circuits that can implement multi-input MAC and convolution computation were constructed by integrating multiplication units in parallel and addition units in series.The construction of multi-functional DNA regulatory circuits has demonstrated the modularity and reliability of these DNA computing units,which hold promise for the construction of complex and large-scale DNA circuits.(3)Using the chemical reaction network as the programming language,a systematic strategy has been demonstrated for implementing convolutional neural network computation at the molecular level.Once an optimal model has been obtained after training in silico,a computational tool translates all parameters and mathematical functions into DNA regulatory circuit that realize the convolutional neural network computation to recognize patterns up to 8 categories,in which 12×12 molecular patterns were encoded with 144 DNA strands.By connecting logic and convolutional neural network circuit architectures,the systematic molecular implementation of DNA hierarchical neural network has been demonstrated,which can recognize patterns up to32 categories.In conclusion,this thesis focuses on the design and construction of DNA molecular systems that can implement complex information processing tasks.A design strategy of simple molecular regulatory elements has been proposed to realize the design and construction of DNA computing units with cascaded and modular properties.Using the recognition domains of DNA molecular as an interface,functional DNA regulatory circuits can be implemented by combining DNA computing units,which can be extended to convolutional neural network computation.