基于文本挖掘的WUI火灾致灾因子网络构建及特征分析

为预防和减少WUI火灾的发生, 挖掘WUI火灾关键致灾因子, 厘清致灾因子间的作用机制. 本文首先基于本文挖掘技术从WUI火灾事故案例中得到致灾因子, 使用Apriori算法得到致灾因子之间的关联规则. 然后使用复杂网络理论构建WUI火灾致灾因子网络, 计算网络拓扑特征参数, 探析WUI火灾致灾因子网络特征. 最后引入WUI火灾致因链风险度指标, 挖掘出高风险连边, 并提出断链措施. 结果表明: WUI火灾致灾因子网络具有小世界特性, 高温、强风、干旱等对其他致灾因子影响较大. 燃烧废弃物、植物起火、应急响应速度、人为纵火、强风在不同致灾因子转换中具有重要作用, 应加强管控. 网络中风险度最高的边是燃烧废弃物→植物起火, 通过颁布禁止擅自燃烧废弃物等规定, 即可切断该风险链, 实现对WUI火灾的预防和主动控制.     

基于知识图谱的双端知识感知图卷积推荐模型

针对现有基于知识图谱的推荐模型仅从用户或项目一端进行特征提取, 从而缺乏对另一端的特征提取的问题, 提出一种基于知识图谱的双端知识感知图卷积推荐模型. 首先, 对于用户、项目及知识图谱中的实体进行随机初始化表征得到初始特征表示; 接着, 采用基于用户和项目的知识感知注意力机制同时从用户、项目两端在知识图谱中进行特征提取; 其次, 使用图卷积网络采用不同的聚合方式聚合知识图谱传播过程中的特征信息并预测点击率; 最后, 为了验证模型的有效性, 在Last.FM和Book-Crossing两个公开数据集上与4个基线模型进行对比实验. 在Last.FM数据集上, AUC和F1分别比最优的基线模型提升了4.4%、3.8%, ACC提升了1.1%. 在Book-Crossing数据集上, AUC和F1分别提升了1.5%、2.2%, ACC提升了1.4%. 实验结果表明, 本文的模型在AUC、F1和ACC指标上比其他的基线模型具有更好的鲁棒性.     

基于图神经网络的最大化代数连通度算法

随着智能体数量的增加, 多智能体系统中潜在的通信链路数量呈指数级增长. 过多冗余链路的存在给系统带来了大量的能源浪费和维护成本, 而盲目地去除链路又会降低系统的稳定性和安全性. 代数连通度是衡量图连通性的重要指标之一. 然而, 传统的半正定规划(SDP)方法和启发式算法在求解大规模场景下的最大化代数连通度问题时非常耗时. 在本文中, 我们提出了一种监督式的图神经网络模型来优化多智能体系统的代数连通度. 我们将传统的SDP方法应用于小规模任务场景中, 得到足够丰富的训练样本和标签. 在此基础上, 我们训练了一个图神经网络模型, 该模型可用于更大规模的任务场景中. 实验结果表明, 当需要去除15条边时, 我们的模型的平均性能达到了传统SDP方法的98.39%. 此外, 我们的模型计算时间极其有限, 可以推广到实时场景中去.     

Protein–Protein Interaction Prediction for Targeted Protein Degradation

Protein–protein interactions (PPIs) play a fundamental role in various biological functions; thus, detecting PPI sites is essential for understanding diseases and developing new drugs. PPI prediction is of particular relevance for the development of drugs employing targeted protein degradation, as their efficacy relies on the formation of a stable ternary complex involving two proteins. However, experimental methods to detect PPI sites are both costly and time-intensive. In recent years, machine learning-based methods have been developed as screening tools. While they are computationally more efficient than traditional docking methods and thus allow rapid execution, these tools have so far primarily been based on sequence information, and they are therefore limited in their ability to address spatial requirements. In addition, they have to date not been applied to targeted protein degradation. Here, we present a new deep learning architecture based on the concept of graph representation learning that can predict interaction sites and interactions of proteins based on their surface representations. We demonstrate that our model reaches state-of-the-art performance using AUROC scores on the established MaSIF dataset. We furthermore introduce a new dataset with more diverse protein interactions and show that our model generalizes well to this new data. These generalization capabilities allow our model to predict the PPIs relevant for targeted protein degradation, which we show by demonstrating the high accuracy of our model for PPI prediction on the available ternary complex data. Our results suggest that PPI prediction models can be a valuable tool for screening protein pairs while developing new drugs for targeted protein degradation.     

The Role of Polymorphisms at the Interleukin-1, Interleukin-4, GATA-3 and Cyclooxygenase-2 Genes in Non-Surgical Periodontal Therapy

Periodontitis is a multifactorial disease. The aim of this explorative study was to investigate the role of Interleukin-(IL)-1, IL-4, GATA-3 and Cyclooxygenase-(COX)-2 polymorphisms after non-surgical periodontal therapy with adjunctive systemic antibiotics (amoxicillin/metronidazole) and subsequent maintenance in a Caucasian population. Analyses were performed using blood samples from periodontitis patients of a multi-center trial (ClinicalTrials.gov {"type":"clinical-trial","attrs":{"text":"NCT00707369","term_id":"NCT00707369"}}NCT00707369=ABPARO-study). Polymorphisms were analyzed using quantitative real-time PCR. Clinical attachment levels (CAL), percentage of sites showing further attachment loss (PSAL) ≥1.3 mm, bleeding on probing (BOP) and plaque score were assessed. Exploratory statistical analysis was performed. A total of 209 samples were genotyped. Patients carrying heterozygous genotypes and single-nucleotide-polymorphisms (SNP) on the GATA-3-IVS4 +1468 gene locus showed less CAL loss than patients carrying wild type. Heterozygous genotypes and SNPs on the IL-1A-889, IL-1B +3954, IL-4-34, IL-4-590, GATA-3-IVS4 +1468 and COX-2-1195 gene loci did not influence CAL. In multivariate analysis, CAL was lower in patients carrying GATA-3 heterozygous genotypes and SNPs than those carrying wild-types. For the first time, effects of different genotypes were analyzed in periodontitis progression after periodontal therapy and during supportive treatment using systemic antibiotics demonstrating a slight association of GATA-3 gene locus with CAL. This result suggests that GATA-3 genotypes are a contributory but non-essential risk factor for periodontal disease progression.     

Novel Insights into Factor D Inhibition

Complement-mediated diseases or complementopathies, such as Paroxysmal nocturnal hemoglobinuria (PNH), cold agglutinin disease (CAD), and transplant-associated thrombotic microangiopathy (TA-TMA), demand advanced complement diagnostics and therapeutics be adopted in a vast field of medical specialties, such as hematology, transplantation, rheumatology, and nephrology. The miracle of complement inhibitors as “orphan drugs” has dramatically improved morbidity and mortality in patients with otherwise life-threatening complementopathies. Efficacy has been significantly improved by upstream inhibition in patients with PNH. Different molecules may exert diverse characteristics in vitro and in vivo. Further studies remain to show safety and efficacy of upstream inhibition in other complementopathies. In addition, cost and availability issues are major drawbacks of current treatments. Therefore, further developments are warranted to address the unmet clinical needs in the field of complementopathies. This state-of-the-art narrative review aims to delineate novel insights into factor D inhibition as a promising target for complementopathies.     

Production of Lacto-N-biose I Using Crude Extracts of Bifidobacterial Cells

Lacto-N-biose I (LNB) is supposed to represent the bifidus factor in human milk oligosaccharides, and can be practically produced from sucrose and GlcNAc using four bifidobacterial enzymes, 1,3-β-galactosyl-N-acetylhexosamine phosphorylase, sucrose phosphorylase, UDP-glucose-hexose 1-phosphate uridylyltransferase, and UDP-glucose 4-epimerase, recombinantly produced by Escherichia coli. Here the production of LNB by the same enzymatic method without using genetically modified enzymes to consider the use of LNB for a food ingredient was reported. All four enzymes were produced as the intracellular enzymes of Bifidobacterium strains. The mixture of the crude extracts contained all four enzymes, with other enzymes interfering with the LNB production, namely, phosphoglucomutase, fructose 6-phosphate phosphoketolase, and glycogen phosphorylase. The first two interfering enzymes were selectively inactivated by heat treatment at 47 °C for 1 h in the presence of pancreatin, and glycogen phosphorylase was disabled by hydrolyzing its possible acceptor molecules using glucoamylase. Finally, 91 % of GlcNAc was converted into LNB in the 100-mL reaction mixture containing 300 mM GlcNAc.