Remote-controlled nucleic acids for biology and medicine

Prof. Dr. Michael Booth, University College London, United Kingdom

DNA and RNA form the basis for many therapeutic and experimental technologies, including gene editing and silencing, several aspects of nanotechnology, aptamers and their applications, and cell-free gene expression. It would be advantageous to control the function of these technologies, as this would greatly expand their application in biology and medicine by reducing toxic on/off-target effects. The main focus of our research is the generation of remote-controlled nucleic acids under the control of various biologically- and medically-applicable stimuli, including temperature, magnetism, and multiple wavelengths of light. We are also exploring several applications of these nucleic acids, for instance to control communication of synthetic cells with living cells and gene delivery/knockdown. In the future, our universal chemical method for controlling DNA and RNA structure and function may form the basis of controllable therapeutics and new technologies for basic research.

Nonvolatile resistive memory technology for deep neural network hardware applications

Dr. Wooseok Choi, IBM Research Europe, Switzerland

In the era of artificial intelligence (AI), the advance of deep neural network (DNN) algorithms has been driving conventional digital hardware over the edge due to power efficiency limitations. Recently, analog AI hardware consisting of crossbar arrays of resistive memory devices, the memristors, have been drawing massive attention as a promising solution to overcome the inherent bottleneck of the von Neumann architecture. They enable a better mapping of the neural network architecture in hardware, in which the resistive devices represent the weights.

Here, memristor (memory+resistor) technology and a crossbar array platform play crucial roles in realizing the analog AI hardware. The memristor devices store synaptic weights as conductance values, and the crossbar array physically implements a neural network on real hardware. Thanks to the compatible design with the NN structure, the RPU system can accelerate NN inference and deep learning through parallel information processing in the memory.  However, realistic device behaviors must be considered to operate analog AI hardware successfully, such as conductance variability, retention, nonlinear conductance response to applied pulses, endurance upon several cycles of weight updates, etc. This lecture extensively explores the memristor technologies for DNN hardware implementation, including basic working principles of devices and arrays in the system, as well as requirements and challenges from the literature.

Biocompatible DNA hydrogels as mechanically tunable smart materials

Dr. Iliya D. Stoev, Karlsruhe Institute of Technology, Germany

Structure and function of materials are tightly interlinked, but detailed understanding of the relation often requires in-depth mechanical characterisation that ensures optimal and tunable performance. However, such detailed characterisation can prove difficult, e.g. in highly delicate and hard-to-predict complex systems, such as concentrated polyurethane foams and glass-forming vitrimers. The difficulty in these systems stems from the very limited degree of control over the microscopic interactions shaping the phase diagram. Hence, to address outstanding challenges in physics aiming at linking microscopic structure to macroscopic properties, we turn to a new generation of smart, stimuli-responsive materials that form due to highly predictable, hierarchical self-assembly of pre-selected functional building blocks. Biocompatible DNA hydrogels serve as programmable and multi-functional platform of smart biomaterials. The mechanical properties of these self-healing networks are highly sequence-specific since they largely depend on strictly imposed binding rules between the nucleotide bases comprising each oligomer. Watson-Crick base-pairing dictates that in DNA sequences adenine only binds to thymine, and guanine only binds to cytosine. Inspired by the simplicity of this genetic code, we discover a ‘zoo’ of different and elaborate viscoelastic networks (DNA hydrogels) originating from linking higher-order DNA structures.