Search

This paper reviews the novel concept of a controllable variational autoencoder (ControlVAE), discusses its parameter tuning to meet application needs, derives its key analytic properties, and offers useful extensions and applications. ControlVAE is a new variational autoencoder (VAE) framework that combines automatic control theory with the basic VAE to stabilize the KL-divergence of VAE models to a specified value. It leverages a non-linear PI controller, a variant of the proportional-integral-derivative (PID) controller, to dynamically tune the weight of the KL-divergence term in the evidence lower bound (ELBO) using the output KL-divergence as feedback. This allows us to precisely control the KL-divergence to a desired value (set point) that is effective in avoiding posterior collapse and learning disentangled representations. While prior work developed alternative techniques for controlling the KL divergence, we show that our PI controller has better stability properties and thus better convergence, thereby producing better disentangled representations from finite training data. In order to improve the ELBO of ControlVAE over that of the regular VAE, we provide a simplified theoretical analysis to inform the choice of set point for the KL-divergence of ControlVAE. We evaluate the proposed method on three tasks: image generation, language modeling, and disentangled representation learning. The results show that ControlVAE can achieve much better reconstruction quality than the other methods for comparable disentanglement. On the language modeling task, our method can avoid posterior collapse (KL vanishing) and improve the diversity of generated text. Moreover, it can change the optimization trajectory, improving the ELBO and the reconstruction quality for image generation. [ABSTRACT FROM AUTHOR]

This paper reviews the novel concept of a controllable variational autoencoder (ControlVAE), discusses its parameter tuning to meet application needs, derives its key analytic properties, and offers useful extensions and applications. ControlVAE is a new variational autoencoder (VAE) framework that combines automatic control theory with the basic VAE to stabilize the KL-divergence of VAE models to a specified value. It leverages a non-linear PI controller, a variant of the proportional-integral-derivative (PID) controller, to dynamically tune the weight of the KL-divergence term in the evidence lower bound (ELBO) using the output KL-divergence as feedback. This allows us to precisely control the KL-divergence to a desired value (set point) that is effective in avoiding posterior collapse and learning disentangled representations. While prior work developed alternative techniques for controlling the KL divergence, we show that our PI controller has better stability properties and thus better convergence, thereby producing better disentangled representations from finite training data. In order to improve the ELBO of ControlVAE over that of the regular VAE, we provide a simplified theoretical analysis to inform the choice of set point for the KL-divergence of ControlVAE. We evaluate the proposed method on three tasks: image generation, language modeling, and disentangled representation learning. The results show that ControlVAE can achieve much better reconstruction quality than the other methods for comparable disentanglement. On the language modeling task, our method can avoid posterior collapse (KL vanishing) and improve the diversity of generated text. Moreover, it can change the optimization trajectory, improving the ELBO and the reconstruction quality for image generation. [ABSTRACT FROM AUTHOR]

Aerosols can modulate the convective cloud-precipitation system in complex ways, which was rarely investigated in North China during the heavily polluted autumn and winter. In this study, the relationships between aerosol pollution and convective precipitation in the North China Plain (112 to 116°E, 36 to 42°N) during autumn and winter were elucidated by combining the data of Global Precipitation Measurement Dual-Frequency Precipitation Radar (GPM-DPR) and the Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2). A positive correlation was found between the aerosol optical depth (AOD) and convective rain rate (RR). Comparative analyses between polluted and clean conditions were conducted to understand this AOD-RR relationship. Higher droplet concentration with smaller size were noted under the polluted condition, associated with higher latent heat released, implying that the increased aerosols may favor the development of the convective cloud-precipitation system through the "aerosol invigoration effect". Besides, the impacts of ambient meteorological conditions on the AOD-RR relationship were also analyzed. It was found that correlations between AOD and RR were always positive in the studied region, irrespective of precipitable water vapor, lower tropospheric stability, and wind shear. A stronger AOD-RR correlation can be found when the lower tropospheric stability or wind shear becomes weaker. Our work provided evidence on the "aerosol invigoration effect" during the heavily polluted months in North China, which may improve the understanding of the complex interactions between aerosol and precipitation. • Aerosol pollution is positively correlated with convective precipitation during autumn and winter in NCP. • GPM-DPR data and MERRA-2 were combined to understand the aerosol-precipitation relationship. • Strong aerosol-precipitation correlation is related to weak wind shear and lower tropospheric stability in the troposphere. [ABSTRACT FROM AUTHOR]

Rapid bone defect regeneration in osteoporotic conditions remains a significant challenge due to the fragile mechanical stability and pathological microenvironment. The absence of bone matrix is the primary characteristic of these defects, and advanced strategies for treating osteoporotic bone defects focus on remodeling the bone matrix's spatial structure and regulating the microenvironment. While many hydrogels have been developed for bone regeneration, their use in repairing osteoporotic bone defects is constrained by deficiencies in shape-adaptivity, weak osteogenic bioactivity, and lack of physiological mechanical support. Herein, a novel bioactive hydrophilic semi-caged NH 2 -T4 silsesquioxane (NH 2 -T4-POSS) nanocage was developed, which was used to modify γ-polyglutamic acid (γ-PGA) together with dopamine, to give an organic/inorganic hybrid hydrogel PGA-DA&T4 for osteoporotic bone regeneration. The developed PGA-DA&T4 hydrogel possesses favorable injectability, shape-adaptivity, self-healability, and strong antioxidant ability. Benefited from organic/inorganic hybridation and multidimensional molecular interacting mechanism, PGA-DA&T4 exhibites enhanced thermal stability and longer degradation period, unique self-polymerizability, high elasticity, and considerable tissue adhesion ability. In vitro experiments proved that PGA-DA&T4 is biocompatible, and is able to promote cell migration and neovascularization, and possesses favorable immunoregulatory to promote macrophage polarization towards anti-inflammatory M2 phenotype. Furthermore, PGA-DA&T4 has been demonstrated to accelerate osteogenic differentiation and inhibit osteoclastogenesis, thereby promoting the repair of osteoporotic bone defects. Our research successfully developed a novel hybrid γ-PGA hydrogel with therapeutic effects and supplied a promising biomaterial with potential clinical application for repairing osteoporotic bone defects. [Display omitted] • Hybrid γ-PGA-based hydrogels were developed as bioactive materials for osteoporotic bone regeneration. • Bioactive Semi-caged NH 2 -T4-POSS nanoparticles were synthesized. • The NH 2 -T4-POSS nanoparticles were used as crosslinker, toughening agent and osteogenesis immunomodulator. • The hybrid hydrogel PGA-DA&T4 can balance the osteogenic immune micro-environment by releasing NH 2 -T4-POSS and catechols. • PGA-DA&T4 can accelerate osteogenic differentiation and inhibit osteoclastogenesis thus promote osteoporotic bone defects repair. [ABSTRACT FROM AUTHOR]

Wuhan, one of the most developed cities in Central China, has been experiencing frequent heavy haze pollution. To understand the impacts of large-scale synoptic patterns and the local-scale planetary boundary layer (PBL) structures on the PM 2.5 pollution in Wuhan, this study applied an objective approach to the classifying of the daily synoptic patterns in 2017. It then combined observational analyses and meteorology-chemistry coupled simulations to investigate a typical pollution episode at the end of November 2017. The synoptic type associated with the heaviest PM 2.5 pollution in Wuhan was characterized by high pressure to the northwest and low pressure to the northeast at the 850-hPa level, which can support northwesterly prevailing winds towards Wuhan. As a result, the aerosols from the highly polluted northern regions can be transported to Wuhan, leading to a high PM 2.5 concentration. Also, when there was high pressure located to the east/southeast of Wuhan at the 850-hPa level, southerly warm advections could be induced. The warming of upper air can significantly suppress the development of PBL by enhancing thermal stability, favoring the accumulation of aerosols. This study elucidated the multi-scale physical mechanisms underlying the aerosol pollution in Wuhan, and has important implications for the forecasting and the mitigating of pollution. • Impacts of boundary layer structure on aerosol pollution in Wuhan were studied. • Synoptic patternsin Central China were classified and polluted types were identified. • Long-range transport of aerosols from north is critical to the pollution in Wuhan. [ABSTRACT FROM AUTHOR]

To investigate the characteristics of particulate matter with an aerodynamic diameter less than 2.5 μm (PM 2.5) and its chemical compositions in the Beijing-Tianjin-Hebei (BTH) region of China during the novel coronavirus disease (COVID-19) lockdown, the ground-based data of PM 2.5 , trace gases, water-soluble inorganic ions, and organic and elemental carbon were analyzed in three typical cities (Beijing, Tianjin, and Baoding) in the BTH region of China from 5-15 February 2020. The PM 2.5 source apportionment was established by combining the weather research and forecasting model and comprehensive air quality model with extensions (WRF-CAMx). The results showed that the maximum daily PM 2.5 concentration reached the heavy pollution level (>150 μg/m3) in the above three cities. The sum concentration of SO 4 2−, NO 3 − and NH 4 + played a dominant position in PM 2.5 chemical compositions of Beijing, Tianjin, and Baoding; secondary transformation of gaseous pollutants contributed significantly to PM 2.5 generation, and the secondary transformation was enhanced as the increased PM 2.5 concentrations. The results of WRF-CAMx showed obviously inter-transport of PM 2.5 in the BTH region; the contribution of transportation source decreased significantly than previous reports in Beijing, Tianjin, and Baoding during the COVID-19 lockdown; but the contribution of industrial and residential emission sources increased significantly with the increase of PM 2.5 concentration, and industry emission sources contributed the most to PM 2.5 concentrations. Therefore, control policies should be devoted to reducing industrial emissions and regional joint control strategies to mitigate haze pollution. [Display omitted] [ABSTRACT FROM AUTHOR]

• Pt-MoP-MoO 3 /C was obtained via integration of Pt and Mo species on Carbon black (Vulcan-C). • Vulcan-C was employed to promote the uniform dispersion of Pt, MoP and MoO 3. • Pt-MoP-MoO 3 /C exhibits superior electrocatalytic activity for HER, with better durability. • Synergy between electro and photo catalysis greatly enhances Pt catalytic capacity. • The amount of H 2 produced on Pt-MoP-MoO 3 /C under irradiation is 5557.93 mmol h−1 g−1. Although Pt catalysts are still the main current electrocatalyst for HER, the rarity of Pt limits its wider application. Herein, an efficient electro-photo catalyst Pt-MoP-MoO 3 /C was constructed to enhance Pt catalytic capacity, via the uniform integration of Pt nanoparticles and MoP on carbon black (Vulcan-C), in which Vulcan-C was used as a support material, Pt and MoP was used as active sites, and MoO 3 was employed as photosensitive species. Electrocatalytic properties of Pt-MoP-MoO 3 /C toward HER are comparable to those of commercial Pt/C (20%), with superior catalytic stability. Particularly, its catalytic performance is significantly enhanced under irradiation, and the amount of H 2 can reach 5557.93 mmol h−1 g−1, which obviously exceeds the value of commercial Pt/C (20%). Efficient electro-photo catalysis occurring on Pt-MoP-MoO 3 /C under irradiation drastically improves Pt catalytic capacity, which provides a new strategy for enhancing Pt usage and promotes the design of cost-effective Pt-based materials for a fast HER. [Display omitted] [ABSTRACT FROM AUTHOR]

Patients with hepatocellular carcinoma (HCC) display poor prognosis because HCC involves a high rate of metastasis and regrowth. Herein, we present an effective strategy to treat HCC using magnetic hyperthermia therapy (MHT)-enhanced cancer immunotherapy combined with transcatheter arterial embolization (TAE). Uniform liquid metal microspheres (LM MSs) obtained by microfluidic technology with powerful eddy-thermal effects could be used as both MHT and TAE agents for effective cancer therapy. The eddy-thermal effect of LM MSs demonstrated effective MHT, whereas LM MS-induced MHT boosted the immune system, promoted immune cell infiltration, and further stimulated powerful immune responses to suppress the growth of distant tumors, together with immune checkpoint blockade therapy. Furthermore, LM MS-lipiodol dispersion displayed excellent efficacy of the combined MHT-TAE in the orthotopic rabbit liver cancer model. Our work not only highlighted that LM MSs could act as effective MHT agents to achieve MHT-enhanced immunotherapy but also presented the significant promise of combining MHT with TAE for the efficient treatment of large orthotopic liver tumors. [ABSTRACT FROM AUTHOR]

Constructing a unique electrochemical interface to enhance the catalytic capacity of Pt-based catalysts is indispensable for wider application of the hydrogen evolution reaction (HER). Herein, platinum-analogous molybdenum carbide (Mo2C) was combined with a lower content of Pt to construct the Pt/Mo2C (C) heterostructure via a solid-phase method, using ammonium molybdate as the precursor. Vulcan-C served as a support to promote the distribution of the Pt and Mo2C heterostructure, and cooperative effects between Pt and the Mo2C heterostructure contributed to the significantly improved catalytic capacity of Pt. The obtained Pt/Mo2C (C) exhibits superior HER activity and enhanced long-term durability in the acidic medium, with a low overpotential of 38 mV at 10 mA cm−2 and a low Tafel slope of 24 mV dec−1. In particular, a drastically enhanced amount of H2 production can be achieved (6837.28 mmol h−1 g−1). This facile approach not only provides a new pathway for constructing novel heterostructures but also gives an insight into the design of cost-effective Pt-based materials for an efficient HER. [ABSTRACT FROM AUTHOR]

Interleukin 2 (IL2) is the first approved immunotherapeutic agent in cancer treatment. However, high‐dose IL2 administrated through intratumoral injection still spreads all over the body, causing serious systemic toxicity. Herein, an injectable nickel‐alginate hydrogel microsphere (Ni‐ALGMS) to allow effective loading of IL2 and its sustained release after intratumoral administration is reported. In this design, histidine (his)‐tagged IL2 is assembled into the Ni‐ALGMS via the coordination bonds between his‐tag and Ni2+. After injecting IL2‐loaded Ni‐ALGMSs (IL2@Ni‐ALGMSs) into the tumor, IL2 slowly releases over long periods, thereby avoiding the risk of cytokine storm happening in IL2 systemic administration. Applying such IL2@Ni‐ALGMSs for tumor model treatment can significantly increase the tumor infiltration of T lymphocytes, and effectively inhibit tumor growth, especially in combination with immune checkpoint inhibitors. This study presents a novel IL2 sustained‐releasing platform for tumor immunotherapy, which can also be conveniently applied in other cytokines‐based immunotherapies. [ABSTRACT FROM AUTHOR]

• Material boundary layer height Hc was proposed. • Hc was relatively lower than Hθ and Hu , despite a high bias caused by the residual layer, multilayer aerosol structure, or lower inversion. • Hc , Hθ , Hu and Ht differed greatly on clean days. During the pollution process, Hc , Hθ and Ht were almost 700 m. The planetary boundary layer (PBL) height mainly determines the environmental capacity for the diffusion of atmospheric pollutants, and has always been a hot issue in the study of air pollution. However, there still remains great uncertainty, partly because different PBL heights definitions and the PBL heights are obtained by various measurement instruments. Pollutants are the substances emitted, different from the atmospheric background physical properties such as wind, temperature and turbulence flux that always exist even without pollution. It is very important to distinguish PBL heights obtained from wind, temperature, turbulence quantities and the concentration of pollutants. In this paper, we express the PBL heights determined on the above four parameters as Hu, H θ , Ht and Hc respectively, and compare them during a heave haze pollution process in Beijing using observation data and simulation results. The comparison results show that: (1) H θ , namely the inversion layer height, decreased from approximately 1250 m to 450 m from 26 to 30 December, resulting in deteriorating pollution situation. Hc, calculated by lidar and characterizes the maximum depth of vertical diffusion of particulates, also dropped below 500 m, and on the whole, the values of Hc estimated by gradient method and H θ were in good agreement; (2) Generally, Hc was relatively lower than H θ and Hu , despite a high bias caused by the existence of the residual layer, multilayer aerosol structure, or lower inversion; (3) Ht estimated from turbulence quantities simulated by WRF model mainly approximated Hu , H θ and Hc in the daytime during haze pollution, however for the nocturnal boundary layer height in the winter, Ht was seriously underestimated. The averaged PBL heights according to the pollution level showed that Hc , H θ , Hu and Ht differed greatly on clean days, and the maximum PBL height Hu exceeded 1400 m. On clean days, the inversion intensities observed were lower, so the blocking effect of the inversion layer to pollutant diffusion was not strong enough, H θ (886 m) deviated from Hc (1111 m). However, Hc and Ht were very close, approximately 1100 m. The decrease of PBL height led to heavy pollution, Hc , H θ and Ht were almost 700 m. Hu was slightly higher and reduced by about 450 m during heavy pollution. The detailed analyses and comparisons of the PBL height from different variables can help improve the rational application of different methods in the determination of PBL height. [ABSTRACT FROM AUTHOR]

Hyperthermia induced by the alternating magnetic field (AMF) to heat magnetic nanomaterials localized within the tumor has attracted wide attention due to its ability in effective tumor ablation, low side effects, and deep tissue penetration. However, rather strong AMFs are often needed to realize effective magnetic hyperthermia of tumors. Herein, it is discovered that nonmagnetic gallium–indium liquid metal (GaIn LM) above certain sizes prepared by the shearing force could be effectively heated up under AMF with a low field intensity owing to the eddy‐thermal effect, which could be utilized for AMF‐induced hyperthermia treatment of tumors. As demonstrated in experiments using mice bearing subcutaneous 4T1 breast tumors and rats with deeply seated orthotopic liver tumors, local injection of GaIn LM followed by AMF treatment could result in effective ablation of those tumors. With great in vitro biocompatibility, such GaIn LM also shows no appreciable in vivo toxicity to mice within two months. This work thus presents a new type of thermal ablation therapy using nonmagnetic, biocompatible, and injectable LM to ablate tumors under a low‐field‐intensity AMF, with the eddy‐thermal heating mechanism distinctive from that in conventional magnetic hyperthermia therapies using magnetic nanoagents. [ABSTRACT FROM AUTHOR]

Considering the huge cost and long test periods required for new drug development, repurposing drugs that have already been applied in the clinic as new cancer treatment candidates represents an attractive alternative. Disulfiram (DSF) was originally used to treat alcoholism and has proven to have anticancer effects with the coadministration of copper ions (Cu2+). However, the limited water-solubility of DSF and systemic toxicity induced by exogenous Cu2+ hinder its practical application. Herein, we constructed pH-responsive lipid-coated calcium phosphate nanoparticles (LCP NPs) co-loaded with Cu2+ and DSF. After intravenous injection, those nanoparticles with long blood half-life preferentially accumulate in tumors, followed by the degradation of nanoparticles in response to the acidic tumor microenvironment, subsequently releasing Cu2+ and DSF to generate cytotoxic metabolite DTC−Copper complex, bis(diethyldithiocarbamate)–copper (CuET) for tumor treatment. In addition to direct cytotoxicity, the active metabolite CuET could effectively induce immunogenic cell death (ICD) of cancer cells to regulate the immunosuppressive tumor microenvironment, contributing to enhanced immune checkpoint blockade (ICB) therapy in triggering systemic immune responses. This work thus demonstrates the great promises of repurposing the old drug DSF as a new ICD inducer with nano-formulation, to achieve improved synergetic tumor-responsive therapy with low side effects. [ABSTRACT FROM AUTHOR]

The comprehensive air quality model with extensions (CAMx)–high-order decoupled direct method (HDDM) was used to simulate ozone (O 3)-NO x -VOCs sensitivity from June 1 to September 30 in 2019 over the Chengdu metropolitan area. In terms of the diurnal variations of O 3 sensitivities to NO x and VOCs emissions, most cities in the Chengdu metropolitan area, except Ya'an, Suining and Ziyang, show a shift in O 3 sensitivity from NO x -limited to VOC-limited from daytime to nighttime, when 8:00 and 19:00 were the cut-off points for the switch in O 3 -sensitive regimes. The average daily maximum 8-h O 3 sensitivity to regional NO x and VOCs emissions show that the Chengdu metropolitan area is mainly NO x -limited. The O 3 sensitivity to local emissions in urban areas of each city is VOC-limited or in a transition regime, while O 3 sensitivities to overall regional emissions are prone to NO x -limited. Therefore, when proposing O 3 control strategies, it is necessary to consider the different measures for the local and regional emissions. First-order sensitivity coefficients combined with the path integral method (PIM) are used to calculate the contributions of precursors' emissions to O 3 concentrations in different sectors, revealing that mobile source accounted for 31.1–36.9%, followed by industrial source, with contribution of 22.2–30.5% to O 3 concentrations. Finally, five emission reduction scenarios were designed with different NO x /VOC reduction ratios, and O 3 concentration changes with different NO x /VOC reduction ratios were calculated based on O 3 sensitivity coefficients. The results show that the influence of local emission reductions on the changes in O 3 concentrations varies significantly among cities. The decrease in O 3 concentrations reach maximum when the NO x /VOC reduction ratio is 3:1 in most of the cities, except Chengdu and Mianyang, with the highest O 3 decrement occurs at the NO x /VOC reduction ratio of 1:3. • O 3 -sensitive regimes were determined by CAMx-HDDM over the Chengdu metropolitan area. • O 3 sensitivities to local and regional emissions were different. • Mobile source was the largest sectorial O 3 contributor in most cities calculated by the PIM method. • An NO x /VOC reduction ratio of 3:1 resulted in a higher O 3 concentration decrease. [ABSTRACT FROM AUTHOR]