Macrophage Membrane-Coated Nanoparticles Sensitize Glioblastoma to Radiation by Suppressing Proneural–Mesenchymal Transformation in Glioma Stem Cells

Radiotherapy is identified as a crucial treatment for patients with glioblastoma, but recurrence is inevitable. The efficacy of radiotherapy is severely hampered partially due to the tumor evolution. Growing evidence suggests that proneural glioma stem cells can acquire mesenchymal features coupled with increased radioresistance. Thus, a better understanding of mechanisms underlying tumor subclonal evolution may develop new strategies. Herein, data highlighting a positive correlation between the accumulation of macrophage in the glioblastoma microenvironment after irradiation and mesenchymal transdifferentiation in glioblastoma are presented. Mechanistically, elevated production of inflammatory cytokines released by macrophages promotes mesenchymal transition in an NF-κB-dependent manner. Hence, rationally designed macrophage membrane-coated porous mesoporous silica nanoparticles (MMNs) in which therapeutic anti-NF-κB peptides are loaded for enhancing radiotherapy of glioblastoma are constructed. The combination of MMNs and fractionated irradiation results in the blockage of tumor evolution and therapy resistance in glioblastoma-bearing mice. Intriguingly, the macrophage invasion across the blood-brain barrier is inhibited competitively by MMNs, suggesting that these nanoparticles can fundamentally halt the evolution of radioresistant clones. Taken together, the biomimetic MMNs represent a promising strategy that prevents mesenchymal transition and improves therapeutic response to irradiation as well as overall survival in patients with glioblastoma.     

Direct Growth of Uniform Bimetallic Core-Shell or Intermetallic Nanoparticles on Carbon via a Surface-Confinement Strategy for Electrochemical Hydrogen Evolution Reaction

Due to the surface inhomogeneity of the solid supports, direct growth of uniform bimetallic nanoparticles (NPs) with controllable structure and size thereon is particularly challenging. Herein, a surface-confinement strategy is reported to directly prepare ultrafine bimetallic Pt M NPs (MFe, Cu, and Co) with structure of core-shell or intermetallic compounds on an N functionalized carbon support (NC). It is found that the N species of NC support can atomically disperse metal cations of precursors, which largely renders uniform nucleation and growth of bimetallic NPs and fine structure modulation of them. In another regard, metal transfer is confined to a narrow region on NC via N-mediation, hence greatly favoring localized particle growth and formation of ultrafine bimetallic NPs. Remarkably, the ultrafine 3.1 ± 0.7 nm intermetallic Pt₃Fe NPs on NC displayed excellent catalytic activity and durability toward electrochemical hydrogen evolution reaction.     

Electret Molecular Configuration Induced Programmable Photonic Memory for Advanced Logic Operation and Data Encryption

Nonvolatile organic photonic transistor (OPT) memories have attracted widespread attention due to their nondestructive readout, remote controllability, and robust tunability. Developing electrets with similar molecular structures but different memory behaviors and light-responsive features is crucial for light-wavelength-modulated data encryption. However, reported OPT memories have yet to meet this challenge. Here a new electret molecule (“H-PDI”) is developed via reconfiguring the linear perylene diimide molecule (“L-PDI”) to a helical shape. Respectively incorporating H-PDI and L-PDI into the floating gate layer results to H-PDI OPT and L-PDI OPT. Attributing to their remarkably different electronic structures and energy bandgaps, H-PDI OPT and L-PDI OPT preferably respond to 405 and 532 nm light irradiation, respectively. Upon electrical programming, data can be written and stored in both memories with good retention features and a high “1”/“0” state current ratio over 10⁵, though the data can only be erased by light with correct wavelengths, rather than the electrical field. Moreover, data stored in a memory array consisting of both H-PDI OPT and L-PDI OPT can only be read out by correct inputs, and wrong inputs will lead to highly deceptive outputs. This study provides a general design strategy of OPT for advanced data encryption and protection.     

C-Shaped Cartilage Development Using Wharton's Jelly-Derived Hydrogels to Assemble a Highly Biomimetic Neotrachea for use in Circumferential Tracheal Reconstruction

A highly biomimetic neotrachea with C-shaped cartilage rings has promising clinical applications in the treatment of circumferential tracheal defects (CTDs) owing to its structure and physiological function. However, to date, most fabricated tracheal cartilages are O-shaped. In this study, finite element analysis demonstrates C-shaped cartilage rings that exhibit better compliance than O-shaped. Hydrogel is developed using methacryloyl-modified decellularized Wharton's jelly matrix (DWJMA) for the regeneration of C-shaped cartilage rings. This novel hydrogel possesses adjustable physicochemical properties and favorable cytocompatibility. When loaded with chondrocytes, DWJMA hydrogels support the optimal cartilage regeneration both in vitro and in vivo. More importantly, a highly biomimetic neotrachea simultaneously simulating the structural and physiological properties of the normal trachea is regenerated via modular assembly of several individual C-shaped cartilage rings. The results demonstrate the highly biomimetic neotrachea have better patency (88.6 ± 6.1% vs 74.4 ± 9.4%, p < 0.05), improve the survival rate, alleviate weight loss and mucoid impaction, than its O-shaped counterpart when used for the treatment of CTDs in a rabbit model. Therefore, this study proposes a novel hydrogel for the regeneration of C-shaped cartilage and provides new insights into the treatment of CTDs using a highly biomimetic neotrachea with C-shaped cartilage rings.     

Comprehensively Strengthened Metal-Oxygen Bonds for Reversible Anionic Redox Reaction

Introducing anionic redox in layered oxides is an effective approach to breaking the capacity limit of conventional cationic redox. However, the anionic redox reaction generally suffers from excessive oxidation of lattice oxygen to O₂ and O₂ release, resulting in local structural deterioration and rapid capacity/voltage decay. Here, a Na_0.71Li_0.22Al_0.05Mn_0.73O₂ (NLAM) cathode material is developed by introducing Al³⁺ into the transition metal (TM) sites. Thanks to the strong Al–O bonding strength and small Al³⁺ radius, the TMO₂ skeleton and the holistic TM–O bonds in NLAM are comprehensively strengthened, which inhibits the excessive lattice oxygen oxidation. The obtained NLAM exhibits a high reversible capacity of 194.4 mAh g^-1 at 20 mA g^-1 and decent cyclability with 98.6% capacity retention over 200 cycles at 200 mA g⁻¹. In situ characterizations reveal that the NLAM experiences phase transitions with an intermediate OP4 phase during the charge–discharge. Theoretical calculations further confirm that the Al substitution strategy is beneficial for improving the overlap between Mn 3d and O 2p orbitals. This finding sheds light on the design of layered oxide cathodes with highly reversible anionic redox for sodium storage.     

Spidermen Strategy for Stable 24% Efficiency Perovskite Solar Cells

The interface energetics-modification plays an important role in improving the power conversion efficiency (PCE) among the perovskite solar cells (PSCs). Considering the low carrier mobility caused by defects in PSCs, a double-layer modification engineering strategy is adopted to introduce the “spiderman” NOBF₄ (nitrosonium tetrafluoroborate) between tin dioxide (SnO₂ and perovskite layers. NO⁺, as the interfacial bonding layer, can passivate the oxygen vacancy in SnO₂, while BF₄⁻ can optimize the defects in the bulk of perovskite. This conclusion is confirmed by theoretical calculation and transmission electron microscopy (TEM). The synergistic effect of NO⁺ and BF₄⁻ distinctly heightens the carrier extraction efficiency, and the PCE of PSCs is 24.04% with a fill factor (FF) of 82.98% and long-term stability. This study underlines the effectiveness of multifunctional additives in improving interface contact and enhancing PCE of PSCs.     

Solid Solvation Assisted Electrical Doping Conserves High-Performance in 500 nm Active Layer Organic Solar Cells

Organic solar cells (OSCs) process fascinating solution-printing capability to achieve low-cost and large-scale manufacture. However, the rapid power conversion efficiency (PCE) decay with active layer thickness enlargement inhibits the implement of OSCs’ potential advantages. To overcome the bottlenecks of PCE decay in thick active layer OSCs, the electrical doping with componential selectivity in bulk heterojunction (BHJ) film is achieved by introducing a solid solvation additive. Benefiting from the higher exciton splitting efficiency together with the longer drift (L_dr) and diffusion (L_diff) lengths, an OSC with 100 nm BHJ film demonstrates a PCE increment from 16.44% to 18.24% with prolonged dark and illuminated storage stabilities. Applying the solid solvation assisted (SSA) doping method in the OSCs with 500 nm active layer, the PCE significantly increases by 31.9%, from the original value of 11.79% to 15.55%. It further improves to 15.84% in a ternary blend thick-film device, which is the record value to the best of our knowledge. Besides, the SSA doping narrows the PCE gap between the 0.04 and 1 cm² devices. All improvements demonstrate the great potential of SSA doping for OSC commercial manufacture, since it optimizes the photovoltaic performance under all practical conditions of long-term, thick-film, and large-area.     

基于多任务学习的多罪名案件信息联合抽取

面向法律文本的实体关系联合抽取技术对于案情关键信息的智能提取至关重要,是智慧司法领域应用中的重要环节。目前的联合抽取方法虽然已经在特定罪名案件的数据集上取得了较好的效果,但是由于模型在训练时只关注了特定罪名类型文本数据的特点,使得模型的泛化能力有限,在应用到多罪名案件的情况下常常使得模型的效果下降。因此引入多任务学习的方法对多罪名情形下的实体关系联合抽取进行了研究,以涉毒类案件和盗窃类案件两大类罪名的文书数据为基础,构建了一个罪名分类任务作为联合抽取的辅助任务,通过基于特征筛选的动态加权多任务模型同时对两个任务进行学习,在单任务模型的基础上整体F1值提升了2.4个百分点,在涉毒类案件和盗窃类案件上的F1值分别提升了1.6和3.2个百分点。     

结合跨层特征融合与级联检测器的防震锤缺陷检测

目的 防震锤可以减少输电线路的周期性震动以降低线路的疲劳损害,定期对防震锤进行巡检非常必要。针对目前无人机巡检输电线路所得航拍图像背景复杂,而种类较多、形状各异以及特性不一的防震锤在航拍图像中占据像素面积很小,导致防震锤检测过程中出现的检测精度低、无法确定缺陷类型等问题,提出了一种结合跨层特征融合和级联检测器的防震锤缺陷检测方法。方法 本文使用无人机对防震锤部件巡检的航拍图像进行数据扩充建立防震锤缺陷检测数据集,并划分了4种缺陷类型,为研究提供了数据基础。首先,以VGG16（Visual GeometryGroup 16-layer network）为基础对1、3、5层特征进行特征融合得到特征图,平衡了语义信息和空间特征;其次,使用3个级联检测器对目标进行分类定位,减小了交并比（intersection over union,IoU）阈值对网络性能的影响;最后,将非极大值抑制法替换为Soft-NMS（soft non-maximum suppression）算法,去除边界框保留了最佳结果。结果 在自建数据集上验证网络模型对4种防震锤缺陷类型的检测效果,与现有基于深度学习的其他6种先进算法相比,本文算法的平均准确率比Fast R-CNN（fast region-based convolutional network）、Faster R-CNN、YOLOv4（you only look onceversion 4）分别提高了13.5%、3.4%、5.8%,比SSD300（single shot MultiBox detector 300）、YOLOv3、RetinaNet分别提高了9.5%、8.5%、8%。与Faster R-CNN相比,本文方法的误检率降低了5.61%,漏检率降低了3.01%。结论 本文提出的防震锤缺陷检测方法对不同背景、不同光照、不同角度、不同尺度、不同种类和不同缺陷种类的防震锤均有较好的检测结果,不但可以更好地提取防震锤的特征,而且还能提高分类和位置预测网络的定位精度,从而有效提高了防震锤缺陷检测算法的准确率,在满足防震锤巡检工作实际检测要求的同时还具有较好的鲁棒性和有效性。     

跨模态表征与生成技术

多媒体数据持续呈现爆发式增长并显现出异源异构的特性,因此跨模态学习领域研究逐渐引起学术和工业界的关注。跨模态表征与生成是跨模态学习的两大核心基础问题。跨模态表征旨在利用多种模态之间的互补性剔除模态之间的冗余,从而获得更为有效的特征表示;跨模态生成则是基于模态之间的语义一致性,实现不同模态数据形式上的相互转换,有助于提高不同模态间的迁移能力。本文系统地分析了国际与国内近年来跨模态表征与生成领域的重要研究进展,包括传统跨模态表征学习、多模态大模型表示学习、图像到文本的跨模态转换和跨模态图像生成。其中,传统跨模态表征学习探讨了跨模态统一表征和跨模态协同表征,多模态大模型表示学习探讨了基于Transformer的模型研究,图像到文本的跨模态转换探讨了图像视频的语义描述、视频字幕语义分析和视觉问答等领域的发展,跨模态图像生成从不同模态信息的跨模态联合表示方法、图像的跨模态生成技术和基于预训练的特定域图像生成阐述了跨模态生成方面的进展。本文详细综述了上述各个子领域研究的挑战性,对比了国内外研究方面的进展情况,梳理了发展脉络和学术研究的前沿动态。最后,根据上述分析展望了跨模态表征与生成的发展趋势和突破口。