Quantifying Atomically Dispersed Catalysts Using Deep Learning Assisted Microscopy.

The catalytic performance of atomically dispersed catalysts (ADCs) is greatly influenced by their atomic configurations, such as atom-atom distances, clustering of atoms into dimers and trimers, and their distributions. Scanning transmission electron microscopy (STEM) is a powerful technique for imaging ADCs at the atomic scale; however, most STEM analyses of ADCs thus far have relied on human labeling, making it difficult to analyze large data sets. Here, we introduce a convolutional neural network (CNN)-based algorithm capable of quantifying the spatial arrangement of different adatom configurations. The algorithm was tested on different ADCs with varying support crystallinity and homogeneity. Results show that our algorithm can accurately identify atom positions and effectively analyze large data sets. This work provides a robust method to overcome a major bottleneck in STEM analysis for ADC catalyst research. We highlight the potential of this method to serve as an on-the-fly analysis tool for catalysts in future in situ microscopy experiments.

Development of carboxymethyl chitosan-coated zein/soy lecithin nanoparticles for the delivery of resveratrol.

The objective of this work is to formulate a zein-based nanocomposite for the delivery of natural polyphenols. A proprietary atomizing/antisolvent precipitation (AAP) process was used to prepare carboxymethyl chitosan (CMC)-coated zein/soy lecithin (SL) nanoparticles (ZLC NPs). At a suitable mass ratio of zein/SL/CMC (100 : 30 : 30), ZLC NPs with desirable redispersibility and physicochemical stability were successfully fabricated. After that, resveratrol (Res) as the representative natural polyphenol was encapsulated in ZLC NPs. The optimized Res/ZLC NPs exhibited a spherical morphology, small size (259.43 ± 2.47 nm), large zeta potential (-47.7 ± 0.66 mV), and high encapsulation efficiency (91.32 ± 4.01%) and loading capacity (5.27 ± 0.35%). Further characterization indicated that Res was encapsulated in the hydrophobic core of the ZLC matrix in an amorphous state. Compared to free Res, Res/ZLC NPs showed a 2.55-fold increase in the Res dissolution rate, a 2.27-fold increase in bioaccessibility, and a 1.69-fold increase in ABTS˙+ scavenging activity. Also, Res/ZLC NPs showed a higher Res retention rate (>68.0%) than free Res (<35.0%) over 45 days of storage. Therefore, ZLC NPs have promising potential as vehicles for natural polyphenols.

A One-Stage Domain Adaptation Network With Image Alignment for Unsupervised Nighttime Semantic Segmentation.

In this paper, we tackle the problem of semantic segmentation for nighttime images that plays an equally important role as that for daytime images in autonomous driving, but is also much more challenging due to very poor illuminations and scarce annotated datasets. It can be treated as an unsupervised domain adaptation (UDA) problem, i.e., applying other labeled dataset taken in the daytime to guide the network training meanwhile reducing the domain shift, so that the trained model can generalize well to the desired domain of nighttime images. However, current general-purpose UDA approaches are insufficient to address the significant appearance difference between the day and night domains. To overcome such a large domain gap, we propose a novel domain adaptation network "DANIA" for nighttime semantic image segmentation by leveraging a labeled daytime dataset (the source domain) and an unlabeled dataset that contains coarsely aligned day-night image pairs (the target daytime and nighttime domains). These three domains are used to perform a multi-target adaptation via adversarial training in the network. Specifically, for the unlabeled day-night image pairs, we use the pixel-level predictions of static object categories on a daytime image as a pseudo supervision to segment its counterpart nighttime image. We also include a step of image alignment to relieve the inaccuracy caused by the misalignment between day-night image pairs by estimating a flow to refine the pseudo supervision produced by daytime images. Finally, a re-weighting strategy is applied to further improve the predictions, especially boosting the prediction accuracy of small objects. The proposed DANIA is a one-stage adaptation framework for nighttime semantic segmentation, which does not train additional day-night image transfer models as a separate pre-processing stage. Extensive experiments on Dark Zurich and Nighttime Driving datasets show that our DANIA achieves state-of-the-art performance for nighttime semantic segmentation.

Rational Fabrication of Folate-Conjugated Zein/Soy Lecithin/Carboxymethyl Chitosan Core-Shell Nanoparticles for Delivery of Docetaxel.

The objective of this work is to design and fabricate a natural zein-based nanocomposite with core-shell structure for the delivery of anticancer drugs. As for the design, folate-conjugated zein (Fa-zein) was synthesized as the inner hydrophobic core; soy lecithin (SL) and carboxymethyl chitosan (CMC) were selected as coating components to form an outer shell. As for fabrication, a novel and appropriate atomizing/antisolvent precipitation process was established. The results indicated that Fa-zein/SL/CMC core-shell nanoparticles (FZLC NPs) were successfully produced at a suitable mass ratio of Fa-zein/SL/CMC (100:30:10) and the freeze-dried FZLC powder showed a perfect redispersibility and stability in water. After that, docetaxel (DTX) as a model drug was encapsulated into FZLC NPs at different mass ratios of DTX to FZLC (MR). When MR = 1:15, DTX/FZLC NPs were obtained with high encapsulation efficiency (79.22 ± 0.37%), small particle size (206.9 ± 48.73 nm), and high zeta potential (-41.8 ± 3.97 mV). DTX was dispersed in the inner core of the FZLC matrix in an amorphous state. The results proved that DTX/FZLC NPs could increase the DTX dissolution, sustain the DTX release, and enhance the DTX cytotoxicity significantly. The present study provides insight into the formation of zein-based complex nanocarriers for the delivery of anticancer drugs.

Development of nimesulide amorphous solid dispersions via supercritical anti-solvent process for dissolution enhancement.

Formulating amorphous solid dispersions (ASDs) is one of the most promising strategies to overcome solubility limitations in drug development. In this work, development of nimesulide (NIM) ASDs via supercritical anti-solvent (SAS) process was proposed, where the mixtures of dichloromethane (DCM) and methanol (MeOH) were selected as the liquid solvent, and the mixtures of hydroxypropyl methylcellulose (HPMC) and polyvinylpyrrolidone (PVP) were the dispersing materials. The effects of NIM/HPMC/PVP (w/w/w) ratio and DCM/MeOH (v/v) ratio on particle solid-state properties were investigated to identify successful operating conditions. NIM-ASDs powders were formed by well separated spherical microparticles, where NIM crystals had transformed into amorphous state completely; the production yield was 93.6 ± 1.14%; and the reproducibility was very high. For NIM-ASDs, intermolecular interactions between NIM and dispersing materials were formed; the residual solvent was far below the ICH limit; and the chemical structure of NIM did not be degraded or disrupted. Moreover, NIM-ASDs increased the NIM solubility in PBS (pH=6.8) more than 5-folds; the dissolution of NIM from NIM-ASDs granules was faster and more complete than that from commercial Aulin® granules in PBS (pH=6.8). Also, NIM-ASDs well hindered the aging in the recrystallization of amorphous NIM during 12-month sealed storage. Overall, development of NIM-ASDs via SAS process presents an opportunity that as a modified product to increase the efficacy of NIM.

Comparison Study on the Prediction of Multiple Molecular Properties by Various Neural Networks.

Various neural networks, including a single layer neural network (SLNN), a deep neural network (DNN) with multilayers, and a convolution neural network (CNN) have been developed and investigated to predict multiple molecular properties simultaneously. The data set of this work contains∼134 kilo molecules and their 15 properties (including rotational constant A, B, and C, dipole moment, isotropic polarizability, energy of HOMO, energy of LUMO, HOMO-LUMO gap energy, electronic spatial extent, zero point vibrational energy, internal energy at 0 K, internal energy at 298.15 K, enthalpy at 298.15 K, free energy at 298.15 K, and heat capacity at 298.15 K) at the hybrid density functional theory (DFT) level from the QM9 database. Coulomb matrix (CM) converted from the database representing every molecule uniquely and its eigenvalue are respectively used as the input of machine learning. The accuracies of predictions have been compared among SLNN, DNN and CNN by analyzing their mean absolute errors (MAEs). Using eigenvalues as input, both SLNN and DNN can give higher accuracy for the prediction of specific energy properties ( U0, U, H, and G). For the prediction of all 15 molecular properties at a time, DNN with a 3-layers network exhibits the best results using the full CM as input. The number of layers in DNN play a key role in the prediction of multiple molecular properties simultaneously. This work may provide possibility and guidance for the selection of different neural networks and input data forms for prediction and validation of multiple parameters according to different needs.