Complex-valued neural networks have many advantages over their real-valued counterparts. Conventional digital electronic computing platforms are incapable of executing truly complex-valued representations and operations. In contrast, optical computing platforms that encode information in both phase and magnitude can execute complex arithmetic by optical interference, offering significantly enhanced computational speed and energy efficiency. However, to date, most demonstrations of optical neural networks still only utilize conventional real-valued frameworks that are designed for digital computers, forfeiting many of the advantages of optical computing such as efficient complex-valued operations. In this article, we highlight an optical neural chip (ONC) that implements truly complex-valued neural networks. We benchmark the performance of our complex-valued ONC in four settings: simple Boolean tasks, species classification of an Iris dataset, classifying nonlinear datasets (Circle and Spiral), and handwriting recognition. Strong learning capabilities (i.e., high accuracy, fast convergence and the capability to construct nonlinear decision boundaries) are achieved by our complex-valued ONC compared to its real-valued counterpart.

Particle trapping and binding in optical potential wells provide a versatile platform for various biomedical applications. However, implementation systems to study multi-particle contact interactions in an optical lattice remain rare. By configuring an optofluidic lattice, we demonstrate the precise control of particle interactions and functions such as controlling aggregation and multi-hopping. The mean residence time of a single particle is found considerably reduced from 7 s, as predicted by Kramer’s theory, to 0.6 s, owing to the mechanical interactions among aggregated particles. The optofluidic lattice also enables single-bacteria-level screening of biological binding agents such as antibodies through particle-enabled bacteria hopping. The binding efficiency of antibodies could be determined directly, selectively, quantitatively and efficiently. This work enriches the fundamental mechanisms of particle kinetics and offers new possibilities for probing and utilising unprecedented biomolecule interactions at single-bacteria level.

No abstract

A (GGGGCC) hexanucleotide repeat (HR) expansion in the C9ORF72 gene, and its associated antisense (CCCCGG) expansion, are considered the major cause behind frontotemporal dementia and amyotrophic lateral sclerosis. We have performed molecular dynamics simulations to characterize the conformation and dynamics of the 12 duplexes that result from the three different reading frames in sense and antisense HRs for both DNA and RNA. These duplexes display atypical structures relevant not only for a molecular level understanding of these diseases but also for enlarging the repertoire of nucleic-acid structural motifs. G-rich helices share common features. The inner G-G mismatches stay inside the helix in Gsyn-Ganti conformations and form two hydrogen bonds (HBs) between the Watson–Crick edge of Ganti and the Hoogsteen edge of Gsyn. In addition, Gsyn in RNA forms a base-phosphate HB. Inner G-G mismatches cause local unwinding of the helix. G-rich double helices are more stable than C-rich helices due to better stacking and HBs of G-G mismatches. C-rich helix conformations vary wildly. C mismatches flip out of the helix in DNA but not in RNA. Least (most) stable C-rich RNA and DNA helices have single (double) mismatches separated by two (four) Watson–Crick basepairs. The most stable DNA structure displays an “e-motif” where mismatched bases flip toward the minor groove and point in the 5′ direction. There are two RNA conformations, where the orientation and HB pattern of the mismatches is coupled to bending of the helix.

No abstract