SHF: Small: DNA Circuits for Analog Computations

Duke University, Durham NC

Investigators

Abstract

Analog devices have potential advantages over Boolean circuits, particularly for performing numerical computations, and analog circuits are often much more compact and require less resources. These advances are enhanced at molecular scales, where resources are scarce and compact designs are crucial. PI proposes extension of DNA computation from Boolean to analog computation. The analog DNA circuits can be used to control a wide variety of molecular devices. The central goals of this project are (i) to develop (design, simulate, and experimentally test) two architectures for analog DNA circuits, (ii) develop DNA-based methods for digital-to-analog and analog-to-digital conversions to allow hybrid analog-digital DNA circuits, and (iii) to provide demonstrations of applications of analog DNA circuits. The work will involve students at all levels: the graduates students, undergraduates, and high school students. Females and minority students will especially be recruited. Students working on this project will receive training at Duke Univ. in computer science, chemistry, and DNA-based nanoscience. In addition, there are also opportunities for summer internships for undergraduates and high school students. Analog DNA circuits have many important potential applications such as analog control devices, where real values are sensed and analog computations provide controlling output. Prior devices for control of chemical reactions systems that provide for molecular species sensing and response have been limited to finite-state control; analog DNA circuits will allow much more sophisticated analog processing and control. DNA-based molecular robotics have allowed devices to operate autonomously (e.g., to walk on a nanostructure) but have been limited to finite-state control, and analog DNA circuits will allow molecular robotics to include real-time analog control circuits to provide much more sophisticated control, e.g. for control articulated joints of a molecular robot?s limb. Many systems that dynamically learn (e.g., neural networks and probabilistic inference) require analog computation, and analog DNA circuits can be used for back-propagation computation of neural nets and Bayesian inference computation of probabilistic inference systems. The project introduces two architectures for molecular-scale analog computation. In both, the input and outputs of analog gates are directly encoded by relative concentrations of input and output strands respectively, without requiring thresholds for converting to Boolean signals. The 1st architecture has 3 gates: addition, subtraction, and multiplication. Analog circuits constructed from these gates can compute polynomials as well as approximate inverse, and division. The 2nd proposed architecture provides a novel DNA-based method to compute analytic functions such as sqrt(x), ln(x), and exp(x) using multiple DNA-based autocatalytic reaction systems working together. The project also introduces DNA analog-to-digital (A/D) and digital-to-analog (D/A) converters that enable the communication between analog and digital DNA circuits. The project includes full-scale designs, simulations, and experimental demonstrations of the two architectures, demonstrations of hybrid analog-digital DNA circuits, and a small-scale demonstration of an application of analog DNA circuits for control of a chemical reaction system: sensing input concentrations of molecules and controlling output of concentrations of molecules.