Deep learning in retrosynthesis planning: datasets, models and tools.

In recent years, synthesizing drugs powered by artificial intelligence has brought great convenience to society. Since retrosynthetic analysis occupies an essential position in synthetic chemistry, it has received broad attention from researchers. In this review, we comprehensively summarize the development process of retrosynthesis in the context of deep learning. This review covers all aspects of retrosynthesis, including datasets, models and tools. Specifically, we report representative models from academia, in addition to a detailed description of the available and stable platforms in the industry. We also discuss the disadvantages of the existing models and provide potential future trends, so that more abecedarians will quickly understand and participate in the family of retrosynthesis planning.

A Novel Digital Closed Loop MEMS Accelerometer Utilizing a Charge Pump.

This paper presents a novel digital closed loop microelectromechanical system (MEMS) accelerometer with the architecture and experimental evaluation. The complicated timing diagram or complex power supply in published articles are circumvented by using a charge pump system of adjustable output voltage fabricated in a 2P4M 0.35 µm complementary metal-oxide semiconductor (CMOS) process, therefore making it possible for interface circuits of MEMS accelerometers to be integrated on a single die on a large scale. The output bitstream of the sigma delta modulator is boosted by the charge pump system and then applied on the feedback comb fingers to form electrostatic forces so that the MEMS accelerometer can operate in a closed loop state. Test results agree with the theoretical formula nicely. The nonlinearity of the accelerometer within ±1 g is 0.222% and the long-term stability is about 774 µg.

Dual-View Learning Based on Images and Sequences for Molecular Property Prediction.

The prediction of molecular properties remains a challenging task in the field of drug design and development. Recently, there has been a growing interest in the analysis of biological images. Molecular images, as a novel representation, have proven to be competitive, yet they lack explicit information and detailed semantic richness. Conversely, semantic information in SMILES sequences is explicit but lacks spatial structural details. Therefore, in this study, we focus on and explore the relationship between these two types of representations, proposing a novel multimodal architecture named ISMol. ISMol relies on a cross-attention mechanism to extract information representations of molecules from both images and SMILES strings, thereby predicting molecular properties. Evaluation results on 14 small molecule ADMET datasets indicate that ISMol outperforms machine learning (ML) and deep learning (DL) models based on single-modal representations. In addition, we analyze our method through a large number of experiments to test the superiority, interpretability and generalizability of the method. In summary, ISMol offers a powerful deep learning toolbox for drug discovery in a variety of molecular properties.

Bidirectional Electromagnetically Induced Transparency Based on Coupling of Magnetic Dipole Modes in Amorphous Silicon Metasurface.

A bidirectional electromagnetically induced transparency (EIT) arising from coupling of magnetic dipole modes is demonstrated numerically and experimentally based on nanoscale a-Si cuboid-bar metasurface. Analyzed by the finite-difference time-domain (FDTD) Solutions, both the bright and dark magnetic dipole mode is excited in the cuboid, while only the dark magnetic dipole mode is excited in the bar. By breaking the symmetry of the cuboid-bar structure, the destructive interference between bright and dark magnetic dipole modes is induced, resulting in the bidirectional EIT phenomenon. The position and amplitude of simulated EIT peak is adjusted by the vertical spacing and horizontal spacing. The EIT metasurface was fabricated by Electron-Beam Lithography and deep silicon etching technique on the a-Si film deposited by Plasma-Enhanced Chemical Vapor Deposition. Measured by a convergent spectrometer, the fabricated sample achieved a bidirectional EIT peak with transmission up to 65% and 63% under forward and backward incidence, respectively. Due to the enhanced magnetic field induced by the magnetic dipole resonance, the fabricated bidirectional EIT metasurface provides a potential way for magnetic sensing and magnetic nonlinearity.

A Plasmonic Infrared Multiple-Channel Filter Based on Gold Composite Nanocavities Metasurface.

A plasmonic near-infrared multiple-channel filter is numerically and experimentally investigated based on a gold periodic composite nanocavities metasurface. By the interference among different excited plasmonic modes on the metasurface, the multipeak extraordinary optical transmission (EOT) phenomenon is induced and utilized to realize multiple-channel filtering. Investigated from the simulated transmission spectrum of the metasurface, the positions and intensity of transmission peaks are tuned by the geometrical parameters of the metasurface and environmental refractive index. The fabricated metasurface approached transmission peaks at 1128 nm, 1245 nm, and 1362 nm, functioning as a three-passbands filter. With advantages of brief single-layer fabrication and multi-frequency selectivity, the proposed plasmonic filter has potential possibilities of integration in nano-photonic switching, detecting and biological sensing systems.