Search

Federated learning (FL) enables distributed clients to collaboratively train a machine learning model without sharing raw data with each other. However, it suffers from the leakage of private information from uploading models. In addition, as the model size grows, the training latency increases due to the limited transmission bandwidth and model performance degradation while using differential privacy (DP) protection. In this paper, we propose a gradient sparsification empowered FL framework with DP over wireless channels, to improve training efficiency without sacrificing convergence performance. Specifically, we first design a random sparsification algorithm to retain a fraction of the gradient elements in each client’s local model, thereby mitigating the performance degradation induced by DP and reducing the number of transmission parameters over wireless channels. Then, we analyze the convergence bound of the proposed algorithm, by modeling a non-convex FL problem. Next, we formulate a time-sequential stochastic optimization problem for minimizing the developed convergence bound, under the constraints of transmit power, the average transmitting delay, as well as the client’s DP requirement. Utilizing the Lyapunov drift-plus-penalty framework, we develop an analytical solution to the optimization problem. Extensive experiments have been implemented on three real-life datasets to demonstrate the effectiveness of our proposed algorithm. We show that our proposed algorithms can fully exploit the interworking between communication and computation to outperform the baselines, i.e., random scheduling, round robin, and delay-minimization algorithms. [ABSTRACT FROM AUTHOR]

Summary: Previously observational studies did not draw a clear conclusion on the association between fatty liver diseases and bone mineral density (BMD). Our large-scale studies revealed that MAFLD and hepatic steatosis had no causal effect on BMD, while some metabolic factors were correlated with BMD. The findings have important implications for the relationship between fatty liver diseases and BMD, and may help direct the clinical management of MAFLD patients who experience osteoporosis and osteopenia. Purpose: Liver and bone are active endocrine organs with several metabolic functions. However, the link between metabolic dysfunction-associated fatty liver disease (MAFLD) and bone mineral density (BMD) is contradictory. Methods: Using the UK Biobank and National Health and Nutrition Examination Survey (NHANES) dataset, we investigated the association between MAFLD, steatosis, and BMD in the observational analysis. We performed genome-wide association analysis to identify single-nucleotide polymorphisms associated with MAFLD. Large-scale two-sample Mendelian randomization (TSMR) analyses examined the potential causal relationship between MAFLD, hepatic steatosis, or major comorbid metabolic factors, and BMD. Results: After adjusting for demographic factors and body mass index, logistic regression analysis demonstrated a significant association between MAFLD and reduced heel BMD. However, this association disappeared after adjusting for additional metabolic factors. MAFLD was not associated with total body, femur neck, and lumbar BMD in the NHANES dataset. Magnetic resonance imaging-measured steatosis did not show significant associations with reduced total body, femur neck, and lumbar BMD in multivariate analysis. TSMR analyses indicated that MAFLD and hepatic steatosis were not associated with BMD. Among all MAFLD-related comorbid factors, overweight and type 2 diabetes showed a causal relationship with increased BMD, while waist circumference and hyperlipidemia had the opposite effect. Conclusion: No causal effect of MAFLD and hepatic steatosis on BMD was observed in this study, while some metabolic factors were correlated with BMD. This has important implications for understanding the relationship between fatty liver disease and BMD, which may help direct the clinical management of MAFLD patients with osteoporosis. [ABSTRACT FROM AUTHOR]

A hybrid control strategy is proposed for a SS compensated three-phase wireless power transfer (3P-WPT) system, which realizes full zero-voltage switching (ZVS) operation of the converter during the entire charging process of a battery. First, a mathematical equivalent model of 3P-WPT is established by fundamental harmonic approximation, the ZVS operation of the active bridge in the constant current (CC) charging mode is optimized by considering the relationship between the input impedance, operation frequency, and duty cycle. Consequently, an optimal modulation strategy featuring full ZVS operation is proposed for the constant current and constant voltage (CV) charging modes of 3P-WPT. Second, based on the NSGA-II algorithm, a mathematical model for the multi-objective optimization of coils is established, which aims for a high transmission frequency and power density. The parameters of coils are optimized accordingly. Finally, a 2 kW experimental prototype is built. Experimental results verify the effectiveness of the proposed control strategy, where the maximum efficiency is up to 90.26% at a full load. [ABSTRACT FROM AUTHOR]

Real-parameter single objective optimization has been studied for decades. In recent, a new setting is applied in this field based on the consideration that solving difficulty scales exponentially with the increase in dimensionality. Under the new setting, differential evolution (DE) still outstands in performance as before. Meanwhile, a new type of population-based metaheuristic—gaining–sharing knowledge-based algorithm, becomes a dark horse. Furthermore, ensemble of the above two types of algorithm is proposed in the literature. Although such ensemble shows good performance, provided that a more reasonable scheme is used for the communication between the constituent algorithms, better ensemble can be obtained. We believe that the new scheme should be with adaptiveness. In this paper, we propose an adaptive scheme for the communication. According to the scheme, individuals chosen based on fitness and lifetime are exchanged. In fact, in the field of DE, it is rare to consider lifetime of individual. However, lifetime is no less important than fitness in our scheme. In our experiment, our ensemble is compared with seven state-of-the-art algorithms. According to experimental results, our ensemble is comparable to one of the peers and better than the other ones. [ABSTRACT FROM AUTHOR]

Multiple myeloma (MM) is a prevalent hematological malignancy that poses significant challenges for treatment. Dysregulated cholesterol metabolism has been linked to tumorigenesis, disease progression, and therapy resistance. However, the correlation between cholesterol metabolism-related genes (CMGs) and the prognosis of MM remains unclear. Univariate Cox regression analysis and LASSO Cox regression analysis were applied to construct an overall survival-related signature based on the Gene Expression Omnibus database. The signature was validated using three external datasets. Enrichment analysis and immune analysis were performed between two risk groups. Furthermore, an optimal nomogram was established for clinical application, and its performance was assessed by the calibration curve and C-index. A total of 6 CMGs were selected to establish the prognostic signature, including ANXA2, CHKA, NSDHL, PMVK, SCAP and SQLE. The prognostic signature demonstrated good prognostic performance and correlated with several important clinical parameters, including number of transplants, International Staging System, albumin, beta2-Microglobulin and lactate dehydrogenase levels. The function analysis and immune analysis revealed that the metabolic pathways and immunologic status were associated with risk score. The nomogram incorporating the signature along with other clinical characteristics was constructed and the discrimination was verified by the calibration curve and C-index. Our findings indicated the potential prognostic connotation of cholesterol metabolism in MM. The development and validation of the prognostic signature is expected to aid in predicting prognosis and guiding precision treatment for MM. [ABSTRACT FROM AUTHOR]

Background: There is still no specific real-world data regarding the clinical activity of immune checkpoint inhibitors in the elderly with liver cancer. Our study aimed to compare the efficacy and safety of immune checkpoint inhibitors between patients aged ≥ 65 years and the younger group, while exploring their differences in genomic background and tumor microenvironment. Methods: This retrospective study was conducted at two hospitals in China and included 540 patients treated with immune checkpoint inhibitors for primary liver cancer between January 2018 and December 2021. Patients' medical records were reviewed for clinical and radiological data and oncologic outcomes. The genomic and clinical data of patients with primary liver cancer were extracted and analyzed from TCGA-LIHC, GSE14520, and GSE140901 datasets. Results: Ninety-two patients were classified as elderly and showed better progression-free survival (P = 0.027) and disease control rate (P = 0.014). No difference was observed in overall survival (P = 0.69) or objective response rate (P = 0.423) between the two age groups. No significant difference was reported concerning the number (P = 0.824) and severity (P = 0.421) of adverse events. The enrichment analyses indicated that the elderly group was linked to lower expression of oncogenic pathways, such as PI3K-Akt, Wnt, and IL-17. The elderly had a higher tumor mutation burden than younger patients. Conclusions: Our results indicated that immune checkpoint inhibitors might exhibit better efficacy in the elderly with primary liver cancer, with no increased adverse events. Differences in genomic characteristics and tumor mutation burden may partially explain these results. [ABSTRACT FROM AUTHOR]

This paper numerically investigates the aerodynamic performance of the T106A low-pressure turbine based with different inflow conditions at moderate Reynolds number by using high performance computing based on high order unstructured methods. Two different inflow conditions respectively of uniform and disturbed are considered, while for the latter a small circular cylinder is placed upstream of the cascade to generate wake turbulence as a long-standing disturbance. A high order Fourier-spectral/hp element method is employed to solve the flow dynamics in the cascade of high complex geometries. Flow transition characteristics are quantified in terms of the distribution of cascade wall surface pressure and friction coefficient, the distribution of wake profile pressure loss and the evolution characteristics of boundary layer flow structures as well. The numerical results show that the current numerical simulations accurately predict the flow transition performance of low-pressure turbine cascades and capture the effects of wake-generated disturbance on the cascade, which is shown to effectively modify the flow transition performance as compared with the uniform inflow case. [ABSTRACT FROM AUTHOR]

Background: Metastasis is the major cause of treatment failure and cancer-related deaths in prostate cancer (PCa) patients. Our previous study demonstrated that a CD44+ subpopulation isolated from PCa cells or tumours possesses both stem cell properties and metastatic potential, serving as metastatic prostate cancer stem cells (mPCSCs) in PCa metastasis. However, the underlying mechanisms remain unknown. Methods: In this study, we established PCa models via the orthotopic and subcutaneous implantation of different human PCa cancer cell lines, and compared the metastatic efficacy, after which process function analysis of target genes was pinpointed. Results: Several novel differentially expressed genes (DEGs) between orthotopic and ectopic tumours were identified. Among them, human homeobox B9 (HOXB9) transcription factor was found to be essential for PCa metastasis, as evidenced by the diminished number of lung metastatic foci derived from orthotopic implantation with HOXB9-deficient CWR22 cells, compared with the control. In addition, HOXB9 protein expression was upregulated in PCa tissues, compared with paracancer and benign prostate hyperplasia tissues. It was also positively correlated with Gleason scores. Gain- and loss-of-function assays showed that HOXB9 altered the expression of various tumour metastasis- and cancer stem cell (CSC) growth-related genes in a transforming growth factor beta (TGFβ)-dependent manner. Moreover, HOXB9 was overexpressed in an ALDH+CD44+CXCR4+CD24+ subpopulation of PCa cells that exhibited enhanced TGFβ-dependent tumorigenic and metastatic abilities, compared with other isogenic PCa cells. This suggests that HOXB9 may contribute to PCa tumorigenesis and metastasis via TGFβ signalling. Of note, ALDH+CD44+CXCR4+CD24+-PCa cells exhibited resistance to castration and antiandrogen therapy and were present in human PCa tissues. Conclusion: Taken together, our study identified HOXB9 as a critical regulator of metastatic mPCSC behaviour. This occurs through altering the expression of a panel of CSC growth- and invasion/metastasis-related genes via TGFβ signalling. Thus, targeting HOXB9 is a potential novel therapeutic PCa treatment strategy. [ABSTRACT FROM AUTHOR]

Correction to: Planta (2022) 256:72 https://doi.org/10.1007/s00425-022-03965-5 In the original publication, fourth author name was incorrectly published as Hongbao Zhu. Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Correction to: Insights into sweet potato SR proteins: from evolution to species-specific expression and alternative splicing. [Extracted from the article]

Affected by limited computing resources and energy, intelligent terminal devices in edge computing systems cannot perform computationally intensive mining tasks in blockchains based on the PoW (proof-of-work) protocol. Therefore, rational terminal devices, as miners, choose to offload mining tasks to other devices or edge computing servers. Aiming at the problem that lightweight devices cannot complete the blockchain mining tasks, this paper firstly proposes a blockchain mining task offloading strategy based on D2D-EC (Device to Device Communication Assisted Edge Computing). Miners offload mining tasks to CMN (Collaborative Mining Network) integrated by mining devices or edge computing server. Secondly, the mobility of devices increases the risk of failure in the blockchain consensus process. Therefore, we develop a prediction method based on Lagrange interpolation to predict the track of devices. The mobility prediction of devices enable miners to make rational offloading strategy, that is, offload fewer tasks to devices with strong mobility to reduce consensus failure costs. In this paper, the interaction between miners and resource suppliers is modeled as a two-stage multi-leader multi-follower Stackelberg game to obtain the best resource requests of miners and best pricing of resource suppliers. To find the NE (Nash Equilibrium) of the Stackelberg game, this paper develops a gradient search-based best response distributed algorithm (BRD). Simulation results show that the algorithm can optimize miners’ utilities and suppliers’ profits quickly, and the proposed prediction method can effectively enable miners to optimize allocation of mining tasks. [ABSTRACT FROM AUTHOR]