贝叶斯优化卷积神经网络公共场所异常声识别

针对公共场所异常声的感知和识别问题,提出一种基于贝叶斯优化卷积神经网络的识别方法。提取声信号的Gammatone倒谱系数、倍频程功率谱、短时能量和谱质心,组合成声信号的特征图。构建卷积神经网络作为分类器,利用递增的卷积核设置和池化操作处理不同尺度的特征。基于贝叶斯优化算法优化卷积神经网络的模型参数,对包括火苗噼啪声、婴儿啼哭声、烟花燃放声、玻璃破碎声和警报声的5种公共场所异常声进行识别。该方法的识别结果与基于不同的特征提取和分类器方案得到的识别结果进行比较,结果表明该方法的识别效果优于其他特征提取和分类器方案的识别效果。最后分析了该方法在不同信噪比噪声干扰下的识别结果,验证了该方法的有效性。     

Integrated System Pharmacology Approaches to Elucidate Multi-Target Mechanism of Solanum surattense against Hepatocellular Carcinoma

Hepatocellular carcinoma (HCC) is one of the most common malignant liver tumors with high mortality. Chronic hepatitis B and C viruses, aflatoxins, and alcohol are among the most common causes of hepatocellular carcinoma. The limited reported data and multiple spectra of pathophysiological mechanisms of HCC make it a challenging task and a serious economic burden in health care management. Solanum surattense (S. surattense) is the herbal plant used in many regions of Asia to treat many disorders including various types of cancer. Previous in vitro studies revealed the medicinal importance of S. surattense against hepatocellular carcinoma. However, the exact molecular mechanism of S. surattense against HCC still remains unclear. In vitro and in silico experiments were performed to find the molecular mechanism of S. surattense against HCC. In this study, the network pharmacology approach was used, through which multi-targeted mechanisms of S. surattense were explored against HCC. Active ingredients and potential targets of S. surattense found in HCC were figured out. Furthermore, the molecular docking technique was employed for the validation of the successful activity of bioactive constituents against potential genes of HCC. The present study investigated the active “constituent–target–pathway” networks and determined the tumor necrosis factor (TNF), epidermal growth factor receptor (EGFR), mammalian target of rapamycin (mTOR), Bcl-2-like protein 1(BCL2L1), estrogen receptor (ER), GTPase HRas, hypoxia-inducible factor 1-alpha (HIF1-α), Harvey Rat sarcoma virus, also known as transforming protein p21 (HRAS), and AKT Serine/Threonine Kinase 1 (AKT1), and found that the genes were influenced by active ingredients of S. surattense. In vitro analysis was also performed to check the anti-cancerous activity of S. surattense on human liver cells. The result showed that S. surattense appeared to act on HCC via modulating different molecular functions, many biological processes, and potential targets implicated in 11 different pathways. Furthermore, molecular docking was employed to validate the successful activity of the active compounds against potential targets. The results showed that quercetin was successfully docked to inhibit the potential targets of HCC. This study indicates that active constituents of S. surattense and their therapeutic targets are responsible for their pharmacological activities and possible molecular mechanisms for treating HCC. Lastly, it is concluded that active compounds of S. surattense act on potential genes along with their influencing pathways to give a network analysis in system pharmacology, which has a vital role in the development and utilization of drugs. The current study lays a framework for further experimental research and widens the clinical usage of S. surattense.     

The Design of Rapid Self-Healing Alginate Hydrogel with Dendritic Crosslinking Network

Self-healing alginate hydrogels play important roles in the biological field due to their biocompatibility and ability to recover after cracking. One of the primary targets for researchers in this field is to increase the self-healing speed. Sodium alginate was oxidized, generating aldehyde groups on the chains, which were then crosslinked by poly(amino) amine (PAMAM) via Schiff base reaction. The dendritic structure was introduced to the alginate hydrogel in this work, which was supposed to promote intermolecular interactions and accelerate the self-healing process. Results showed that the hydrogel (ADA-PAMAM) formed a gel within 2.5 min with stable rheological properties. Within 25 min, the hydrogel recovered under room temperature. Furthermore, the aldehyde degree of alginate dialdehyde with a different oxidation degree was characterized through gel permeation chromatograph aligned with multi-angle laser light scattering and ultraviolet (UV) absorption. The chemical structure of the hydrogel was characterized through Fourier transform infrared spectroscopy and UV-vis spectra. The SEM and laser scanning confocal microscope (CLSM) presented the antibiotic ability of ADA-PAMAM against both S. aureus and E. coli when incubated with 10⁻⁷ CFU microorganism under room temperature for 2 h. This work presented a strategy to promote the self-healing of hydrogel through forming a dendritic dynamic crosslinking network.     

Integrated Solid-Phase Extraction,Ultra-High-Performance Liquid Chromatography–Quadrupole-Orbitrap High-Resolution Mass Spectrometry,and Multidimensional Data-Mining Techniques to Unravel the Metabolic Network of Dehydrocostus Lactone in Rats

Dehydrocostus lactone (DL) is among the representative ingredients of traditional Chinese medicine (TCM), with excellent anticancer, antibacterial, and anti-inflammatory activities. In this study, an advanced strategy based on ultra-high-performance liquid chromatography–quadrupole-Orbitrap high-resolution mass spectrometry (UHPLC–Q-Orbitrap HRMS) was integrated to comprehensively explore the metabolic fate of DL in rats. First, prior to data collection, all biological samples (plasma, urine, and feces) were concentrated and purified using solid-phase extraction (SPE) pre-treatment technology. Then, during data collection, in the full-scan (FS) data-dependent acquisition mode, FS-ddMS² was intelligently combined with FS-parent ion list (PIL)-dynamic exclusion (DE) means for targeted monitoring and deeper capture of more low-abundance ions of interest. After data acquisition, data-mining techniques such as high-resolution extracted ion chromatograms (HREICs), multiple mass defect filters (MMDFs), diagnostic product ions (DPIs), and neutral loss fragments (NLFs) were incorporated to extensively screen and profile all the metabolites in multiple dimensions. As a result, a total of 71 metabolites of DL (parent drug included) were positively or tentatively identified. The results suggested that DL in vivo mainly underwent hydration, hydroxylation, dihydrodiolation, sulfonation, methylation, dehydrogenation, dehydration, N-acetylcysteine conjugation, cysteine conjugation, glutathione conjugation, glycine conjugation, taurine conjugation, etc. With these inferences, we successfully mapped the “stepwise radiation” metabolic network of DL in rats, where several drug metabolism clusters (DMCs) were discovered. In conclusion, not only did we provide a refined strategy for inhibiting matrix effects and fully screening major-to-trace metabolites, but also give substantial data reference for mechanism investigation, in vivo distribution visualization, and safety evaluation of DL.     

Metabolites Identification and Mechanism Prediction of Neobavaisoflavone In Vitro and In Vivo of Rats through UHPLC-Q-Exactive Plus Orbitrap MS Integrated Network Pharmacology

Neobavaisoflavone is an important isoflavone component isolated from Psoraleae Fructus. It is used extensively worldwide because of its antibacterial, antioxidant, anti-inflammatory, anticancer, and anti-osteoporotic activities. However, there is no systematic and comprehensive research on the metabolism of neobavaisoflavone in vivo and in vitro. The study aimed to analyze the metabolic characteristics and mechanism of neobavaisoflavone for the first time. Firstly, biological samples were pretreated by the solid-phase extraction (SPE) method, methanol precipitation, and acetonitrile precipitation. Secondly, the samples were analyzed on UHPLC-Q-Exactive Plus Orbitrap MS. Thirdly, metabolites were tentatively identified based on retention time, parallel reaction monitoring strategy, diagnostic product ions, and neutral loss fragments. A total of 72 metabolites of neobavaisoflavone were tentatively identified, including 28 in plasma, 43 in urine, 18 in feces, six in the liver, and four in the liver microsome. The results suggested that neobavaisoflavone mainly underwent glucuronidation, sulfation, hydroxylation, methylation, cyclization, hydration, and their composite reactions in vivo and in vitro. In addition, nine active components with high bioavailability and 191 corresponding targets were predicted by the Swiss Drug Design database. The 1806 items of GO and 183 KEGG signaling pathways were enriched. These results showed that metabolites expanded the potential effects of neobavaisoflavone. The present study would provide the scientific basis for the further exploitation and application of neobavaisoflavone.     

Design of an Internal/External Bicontinuous Conductive Network for High-Performance Asymmetrical Supercapacitors

High-energy density supercapacitors have attracted extensive attention due to their electrode structure design. A synergistic effect related to core–shell structure can improve the energy storage capacity and power density of electrode materials. The Ni-foam (NF) substrate coupled with polypyrrole (PPy) conductive coating can serve as an internal/external bicontinuous conductive network. In this work, the distinctive PPy@FeNi₂S₄@NF and PPy@NiCo₂S₄@NF materials were prepared by a simple two-step hydrothermal synthesis with a subsequent in situ polymerization method. PPy@FeNi₂S₄@NF and PPy@NiCo₂S₄@NF could deliver ultrahigh specific capacitances of 3870.3 and 5771.4 F·g⁻¹ at 1 A·g⁻¹ and marvelous cycling capability performances of 81.39% and 93.02% after 5000 cycles. The asymmetric supercapacitors composed of the prepared materials provided a high-energy density of over 47.2 Wh·kg⁻¹ at 699.9 W·kg⁻¹ power density and 67.11 Wh·kg⁻¹ at 800 W·kg⁻¹ power density. Therefore, the self-assembled core–shell structure can effectively improve the electrochemical performance and will have an effective service in advanced energy-storage devices.     

三维网络结构NiCo2S4@NiCo2O4复合电极的制备及其性能

超级电容器具有更大的功率密度、优秀的循环稳定性、极快的充放电速度、超长的循环寿命以及环境友好等突出特点,其性能与构件关系密切,其中最根本的就是组成它的电极材料。本研究主要采用传统的水热法制备出钴酸镍(NiCo₂O₄)电极材料,进而通过离子交换(二次水热)制得镍钴硫(NiCo₂S₄),最后利用化学浴沉积(CBD)法使其与钴酸镍复合,得到最终所需的三维网络结构NiCo₂S₄@NiCo₂O₄复合电极。经过表面形貌表征、循环伏安测试、恒电流充放电测试以及比电容计算分析等可以证明:三维网络结构NiCo₂S₄@NiCo₂O₄复合电极的比电容及循环稳定性等远远优于复合前单一的纯NiCo₂O₄电极材料,具有极大应用前景。     

Conserved Control Path in Multilayer Networks

The determination of directed control paths in complex networks is important because control paths indicate the structure of the propagation of control signals through edges. A challenging problem is to identify them in complex networked systems characterized by different types of interactions that form multilayer networks. In this study, we describe a graph pattern called the conserved control path, which allows us to model a common control structure among different types of relations. We present a practical conserved control path detection method (CoPath), which is based on a maximum-weighted matching, to determine the paths that play the most consistent roles in controlling signal transmission in multilayer networks. As a pragmatic application, we demonstrate that the control paths detected in a multilayered pan-cancer network are statistically more consistent. Additionally, they lead to the effective identification of drug targets, thereby demonstrating their power in predicting key pathways that influence multiple cancers.     

An Improvement to Conformer-Based Model for High-Accuracy Speech Feature Extraction and Learning

Owing to the loss of effective information and incomplete feature extraction caused by the convolution and pooling operations in a convolution subsampling network, the accuracy and speed of current speech processing architectures based on the conformer model are influenced because the shallow features of speech signals are not completely extracted. To solve these problems, in this study, we researched a method that used a capsule network to improve the accuracy of feature extraction in a conformer-based model, and then, we proposed a new end-to-end model architecture for speech recognition. First, to improve the accuracy of speech feature extraction, a capsule network with a dynamic routing mechanism was introduced into the conformer model; thus, the structural information in speech was preserved, and it was input to the conformer blocks via sequestered vectors; the learning ability of the conformed-based model was significantly enhanced using dynamic weight updating. Second, a residual network was added to the capsule blocks, thus, the mapping ability of our model was improved and the training difficulty was reduced. Furthermore, the bi-transformer model was adopted in the decoding network to promote the consistency of the hypotheses in different directions through bidirectional modeling. Finally, the effectiveness and robustness of the proposed model were verified against different types of recognition models by performing multiple sets of experiments. The experimental results demonstrated that our speech recognition model achieved a lower word error rate without a language model because of the higher accuracy of speech feature extraction and learning using our model architecture with a capsule network. Furthermore, our model architecture benefited from the advantage of the capsule network and the conformer encoder, and also has potential for other speech-related applications.     

基于独立组分分析和BP神经网络的可见/近红外光谱蜂蜜品牌的鉴别

提出了一种基于独立组分分析的可见/近红外光谱透射技术快速鉴别蜂蜜品牌的新方法。用独立组分分析方法获取蜂蜜的可见/近红外光谱载荷图,将载荷图中相关性最大的波段,作为人工神经网络的输入建立蜂蜜品牌的鉴别模型。建立了一个三层的BP神经网络模型,各层传递函数采用S型(Sigmoid)函数,并设置网络输入层节点数为9,隐含层节点数为10,输出层节点数为3。每个品牌25个样本,3个品牌共75个样本,用来建立BP神经网络模型,剩余的3个品牌各５个样本用于预测,鉴别准确率达100%,模型的拟合残差为8.245 365×10-5。说明基于独立组分分析的方法具有很好的鉴别效果,为蜂蜜的品牌鉴别提供了一种新方法。