Search

The oxidative transformation of lignin into aromatic compounds is an attractive route for chemical utilization of lignocellulosic biomass. Unlike hydrogenolysis, no consumption of expensive hydrogen is required for the oxidative transformation. However, only limited success has been achieved for the oxidative conversion of lignin. Here, we report that cerium oxide-supported palladium nanoparticles (Pd/CeO2) can efficiently catalyze the one-pot oxidative conversion of 2-phenoxy-1-phenylethanol, a lignin model compound containing a β-O-4 bond and a Cα-hydroxyl group, in methanol in the presence of O2, producing phenol, acetophenone and methyl benzoate as the major products. Pd nanoparticles played a pivotal role in the oxidation of a Cα-hydroxyl group into a Cα-ketonic group, which was crucial for the transformation of the model compound. The presence of the Cα-ketonic group activated the β-O-4 bond, which was subsequently cleaved over the Pd/CeO2 catalyst, affording phenol and acetophenone. At the same time, the Cα–Cβ bond also underwent oxidative cleavage catalyzed by CeO2, producing benzoic acid and further methyl benzoate. The Pd/CeO2 catalyst could also catalyze the oxidative conversion of organosolv lignin under mild conditions (458 K), producing vanillin, guaiacol and 4-hydroxybenzaldehyde. [ABSTRACT FROM AUTHOR]

Photo quality assessment (PQA) aims at computationally and precisely evaluating the quality of images from the aspect of aesthetic. Image aesthetic is strongly correlated with composition. However, few existing works have taken composition into consideration. Besides, existing composition features are typically hand-crafted. In this paper, we propose a novel end-to-end framework for representation learning of image composition. Specially, we build a fully connected graph based on deep features in Convolutional Neural Networks (CNNs). In the graph, edge attributes i.e. similarities between deep features at different positions are used for representing image composition. Besides, we use global attributes of the graph to represent miscellaneous aesthetic aspects. Finally, we use a gate unit to combine both composition features and miscellaneous aesthetic features for aesthetic prediction. The whole network can be trained in an end-to-end manner. Experimental results show that the proposed techniques significantly improves the prediction precision of aesthetic and composition over various datasets. We have released our codes at:. [ABSTRACT FROM AUTHOR]

A reverse dictionary (RD) is a tool designed to find the appropriate word based on a given definition. While large language models (LLMs) have demonstrated exceptional natural language generation capabilities, offering a promising solution for this task, existing LLMs, particularly those with smaller parameter sizes, still require performance enhancements on RD. Fine-tuning LLMs is a direct method of improvement, but it comes with high computational costs and the potential sacrifice of the LLMs' ability to solve other problems. In contrast, employing prompt engineering to construct high-quality prompts that evoke the latent potential of LLMs on RD is a more ideal approach. To this end, we propose a method of automatically constructing prompts based on a small-scale language model. Our method analyzes definitions across multiple semantic dimensions and introduces negative samples to enhance the model's universality and adaptability. Consequently, our small model can generate a candidate set containing the target word with an accuracy rate of 93.1%. We then use these candidates to form effective prompts for LLMs, significantly improving the performance of larger models. Evaluations on a new dataset from the Oxford dictionary show that our LLMs can directly select the target word with an accuracy of 57.5% , which increases to 73.8% when given five candidate words. Our research not only enhances the performance of LLMs on RD but also provides a new perspective on how small models can augment large models, which is particularly important when computational resources are limited. [ABSTRACT FROM AUTHOR]

A reverse dictionary is a task of finding appropriate words based on given definitions. Existing studies struggle to capture the subtle differences between words with similar definitions. Motivated by humans' ability to verify their understanding of a word's semantics by making example sentences and recognizing parts of speech, we propose a reverse dictionary model based on multitask learning (RDMTL), which can alleviate the problem. RDMTL comprises a primary task component that extracts different levels of semantic features from definitions and two auxiliary task components that predict part-of-speech tags and generate sentences for target words. Through jointly learning these three tasks, RDMTL can enhance the understanding of definitions and discover subtle differences among words with similar meanings. Moreover, it can generate more accurate and natural example sentences. We evaluate RDMTL on a modified version of the New Oxford dataset and compare its performance with several baseline models. The experimental results show that RDMTL enhances the rank value metric for the reverse dictionary task by 1.02%, the F1 value metric for the part-of-speech classification task by 20.47%, and the SB-4 metric for the sentence generation task by 14.7%. In addition, to analyze the contribution of each component and the impact of multitask learning, we conducted an ablation study. In this study, a new method and perspective for the reverse dictionary task is introduced. [ABSTRACT FROM AUTHOR]

Eukaryotic yeasts have a variety of subcellular compartments and are ideal platform strains for the construction of complex heterologous natural product biosynthesis pathways. Improving the synthesis efficiency of microbial cell factories through the utilization and modification of subcellular compartments by synthetic biology has good application prospects. Here, we used the yeast PLN1 protein to target the normally endoplasmic reticulum (ER)-localized cytochrome P450 enzyme protopanaxadiol (PPD) synthase (PPDS) to lipid droplets (LDs), which are the storage organelles of the PPDS substrate dammarenediol-II (DD). The efficiency of converting DD to PPD was significantly increased by 394%, and the conversion rate of DD increased from 17.4% to 86.0%. Furthermore, increasing the volume of LDs can significantly enhance the production of DD and its derivatives, but the change in the ratio of the volume and surface area of LDs decreased the conversion efficiency of DD to PPD. Additionally, the biosynthetic pathways of the PPD-type saponin ginsenoside compound K (CK) was reconstituted in a PPD-producing chassis strain, and CK production reached 21.8 mg/L/OD, 4.4-fold higher compared to the native ER-expression strategy. Next, we enhanced the expression of the Pn3-29 gene module to further reduce the accumulation of PPD and increase the production of CK to 41.3 mg/L/OD. Finally, the CK titer of the resulting strain reached 5 g/L in 5 L fed-batch fermentations. This study provides a new strategy for engineering yeast to produce complex natural products. • A new strategy for engineering yeast subcellular compartments to produce complex natural products. • The yeast PLN1 protein can be used to target enzymes to lipid droplets. • Increasing the local concentrations of the substrate and enzyme can effectively increase the yield of the target product. • The maximal titer of ginsenoside CK reached 5 g/L in 5 L a fed-batch fermentations. [ABSTRACT FROM AUTHOR]

The production of knot-free and high-quality wood can be a vital issue in silviculture and forest management. In this study, knot properties, branch occlusion, and wood discoloration were investigated in an 11-year-old Mytilaria laosensis plantation in Guangxi, China, to examine the effectiveness of natural pruning in reducing knot-related defects. A total of 1513 occluded branches from 20 trees were sampled and dissected. Occluded branches were most common at heights of 2–6 m, and the mean diameter of the occluded branches gradually increased with height from the base to 6 m. Linear and generalized linear mixed-effect models were developed to reveal the relationship between branch occlusion and discoloration. The mortality of branches was highest in the third and fourth year and it took roughly three to six years to occlude. The mixed-models indicated that the branch occlusion time was positively correlated with the occluded branch diameter and dead branch stub length, and negatively correlated with the stem radial increment during branch occlusion. Branch discoloration was positively correlated with the occluded branch diameter and branch occlusion time, and was negatively correlated with the branch angle. The probability of wood discoloration also increased with the branch occlusion time. Our findings contribute to a better understanding of the spatiotemporal allocation and internal characteristics of occluded branches, and provide a reference for growing high-quality M. laosensis wood. [ABSTRACT FROM AUTHOR]