Explore the vast range of titles available.

With the prevalence of pre-trained language models (PLMs) and the pre-training–fine-tuning paradigm, it has been continuously shown that larger models tend to yield better performance. However, as PLMs scale up, fine-tuning and storing all the parameters is prohibitively costly and eventually becomes practically infeasible. This necessitates a new branch of research focusing on the parameter-efficient adaptation of PLMs, which optimizes a small portion of the model parameters while keeping the rest fixed, drastically cutting down computation and storage costs. In general, it demonstrates that large-scale models could be effectively stimulated by the optimization of a few parameters. Despite the various designs, here we discuss and analyse the approaches under a more consistent and accessible term ‘delta-tuning’, where ‘delta’ a mathematical notation often used to denote changes, is borrowed to refer to the portion of parameters that are ‘changed’ during training. We formally describe the problem and propose a unified categorization criterion for existing delta-tuning methods to explore their correlations and differences. We also discuss the theoretical principles underlying the effectiveness of delta-tuning and interpret them from the perspectives of optimization and optimal control. Furthermore, we provide a holistic empirical study on over 100 natural language processing tasks and investigate various aspects of delta-tuning. With comprehensive study and analysis, our research demonstrates the theoretical and practical properties of delta-tuning in the adaptation of PLMs. Training a deep neural network can be costly but training time is reduced when a pre-trained network can be adapted to different use cases. Ideally, only a small number of parameters needs to be changed in this process of fine-tuning, which can then be more easily distributed. In this Analysis, different methods of fine-tuning with only a small number of parameters are compared on a large set of natural language processing tasks.

EEG signals exhibit commonality and variability across subjects, sessions, and tasks. But most existing EEG studies focus on mean group effects (commonality) by averaging signals over trials and subjects. The substantial intra- and inter-subject variability of EEG have often been overlooked. The recently significant technological advances in machine learning, especially deep learning, have brought technological innovations to EEG signal application in many aspects, but there are still great challenges in cross-session, cross-task, and cross-subject EEG decoding. In this work, an EEG-based biometric competition based on a large-scale M3CV (A Multi-subject, Multi-session, and Multi-task Database for investigation of EEG Commonality and Variability) database was launched to better characterize and harness the intra- and inter-subject variability and promote the development of machine learning algorithm in this field. In the M3CV database, EEG signals were recorded from 106 subjects, of which 95 subjects repeated two sessions of the experiments on different days. The whole experiment consisted of 6 paradigms, including resting-state, transient-state sensory, steady-state sensory, cognitive oddball, motor execution, and steady-state sensory with selective attention with 14 types of EEG signals, 120000 epochs. Two learning tasks (identification and verification), performance metrics, and baseline methods were introduced in the competition. In general, the proposed M3CV dataset and the EEG-based biometric competition aim to provide the opportunity to develop advanced machine learning algorithms for achieving an in-depth understanding of the commonality and variability of EEG signals across subjects, sessions, and tasks.

Offshore wind power stands out as a promising renewable energy source, offering substantial potential for achieving low carbon emissions and enhancing energy security. Despite its potential, the expansion of offshore wind power faces considerable constraints in offshore power transmission. Hydrogen production derived from offshore wind power emerges as an efficient solution to overcome these limitations and effectively transport energy. This study systematically devises diverse hydrogen energy supply chains tailored to the demands of the transportation and chemical industries, meticulously assessing the levelized cost of hydrogen (LCOH). Our findings reveal that the most cost-efficient means of transporting hydrogen to the mainland is through pipelines, particularly when the baseline distance is 50 km and the baseline electricity price is 0.05 USD/kWh. Notably, delivering hydrogen directly to the port via pipelines for chemical industries proves considerably more economical than distributing it to hydrogen refueling stations, with a minimal cost of 3.6 USD/kg. Additionally, we assessed the levelized cost of hydrogen (LCOH) for supply chains that transmit electricity to ports via submarine cables before hydrogen production and subsequent distribution to chemical plants. In comparison to offshore hydrogen production routes, these routes exhibit higher costs and reduced competitiveness. Finally, a sensitivity analysis was undertaken to scrutinize the impact of delivery distance and electricity prices on LCOH. The outcomes underscore the acute sensitivity of LCOH to power prices, highlighting the potential for substantial reductions in hydrogen prices through concerted efforts to lower electricity costs.

Background Chronic kidney disease (CKD) is a global public health concern. Therefore, to provide timely intervention for non-hospitalized high-risk patients and rationally allocate limited clinical resources is important to mine the key factors when designing a CKD prediction model. Methods This study included data from 1,358 patients with CKD pathologically confirmed during the period from December 2017 to September 2020 at Zhongshan Hospital. A CKD prediction interpretation framework based on machine learning was proposed. From among 100 variables, 17 were selected for the model construction through a recursive feature elimination with logistic regression feature screening. Several machine learning classifiers, including extreme gradient boosting, gaussian-based naive bayes, a neural network, ridge regression, and linear model logistic regression (LR), were trained, and an ensemble model was developed to predict 24-hour urine protein. The detailed relationship between the risk of CKD progression and these predictors was determined using a global interpretation. A patient-specific analysis was conducted using a local interpretation. Results The results showed that LR achieved the best performance, with an area under the curve (AUC) of 0.850 in a single machine learning model. The ensemble model constructed using the voting integration method further improved the AUC to 0.856. The major predictors of moderate-to-severe severity included lower levels of 25-OH-vitamin, albumin, transferrin in males, and higher levels of cystatin C. Conclusions Compared with the clinical single kidney function evaluation indicators (eGFR, Scr), the machine learning model proposed in this study improved the prediction accuracy of CKD progression by 17.6% and 24.6%, respectively, and the AUC was improved by 0.250 and 0.236, respectively. Our framework can achieve a good predictive interpretation and provide effective clinical decision support.

Primary membranous nephropathy (PMN) is the most frequent cause of nephrotic syndrome in adults. Rituximab monotherapy has emerged as a front-line treatment for patients with PMN, but potential markers for predicting the response to rituximab are unknown. In this single-arm retrospective pilot study, 48 patients with PMN without previous immunosuppressive therapy were enrolled. All patients were treated with rituximab and were followed up for at least 6 months. The primary end point was the achievement of complete or partial remission at 6 months. The subsets of lymphocytes were collected at baseline, 1 month, 3 months and 6 months to identify prognostic factors for achieving remission of PMN with rituximab therapy. A total of 58.3% of patients (28/48) achieved remission. Lower serum creatinine, greater serum albumin, and greater phospholipase A2 receptor antigen detected in kidney biopsy at baseline were found in the remission group. After multiple adjustments, a high percentage of natural killer (NK) cells at baseline, especially ≥15.7%, was strongly associated with remission (relative risk = 1.62; 95% CI, 1.00–2.62; P = 0.049), and patients with a response to rituximab had a greater mean percentage of NK cells during the follow-up period compared with nonresponders. Analysis using a receiver operating characteristic curve indicated prognostic value of the NK-cell percentage at baseline, with an area under the curve of 0.716 (95% CI, 0.556–0.876; P = 0.021). The findings from this retrospective pilot study suggest that a high percentage, especially ≥15.7%, of NK cells at baseline might predict a response to rituximab treatment. These findings provide a basis for designing larger-scale studies to test the predictive value of NK cells in patients with PMN undergoing rituximab treatment.

This study aimed to investigate the mechanisms and specific targets of cordycepin in the treatment of renal fibrosis using a unilateral ischemia-reperfusion (UIR) model. A UIR mouse model was established, followed by intraperitoneal injections of cordycepin and Mdivi-1. Masson's trichrome staining and PAS staining were used to identify renal tubulointerstitial fibrosis and assess the degree of renal injury. Fibrosis markers and mitochondrial dynamics-related proteins were evaluated using Western blotting, while differential gene expression and pathway enrichment were analyzed by RNA-seq. Molecular docking, molecular dynamics simulations and surface plasmon resonance were conducted to validate the specific binding sites of cordycepin on the target protein Drp1. Immunofluorescence and in vitro experiments further elucidated the therapeutic mechanism of cordycepin. In vivo experiments showed that intraperitoneal injection of cordycepin significantly reduced renal inflammation and fibrosis, lowered serum creatinine levels, and decreased collagen deposition. Transcriptome analysis revealed that cordycepin treatment downregulated the mitochondrial fission pathway and upregulated the mitochondrial fusion pathway. Western blotting showed reduced levels of fibrosis markers α-SMA and FN, as well as downregulation of Drp1, MFF, and Fis1, and upregulation of OPA1 and Mfn2. In vitro, cordycepin inhibited TGF-β-induced injury in NRK-52E cells, reducing Drp1 expression and IL-6 secretion. Crosstalk experiments confirmed that decreased IL-6 levels were crucial for cordycepin anti-fibrotic effects by suppressing fibroblast activation. Cordycepin ameliorates renal fibrosis by targeting Drp1 to inhibit mitochondrial fission in injured renal tubular epithelial cells, reducing IL-6 secretion and inhibiting fibroblast activation.

Background Disruptions in intracellular pH (pH i ) homeostasis, causing deviations from the physiological range, can damage renal epithelial cells. However, the existence of an adaptive mechanism to restore pH i to normalcy remains unclear. Early research identified H + as a critical mediator of ischemic preconditioning (IPC), leading to the concept of acidic preconditioning (AP). This concept proposes that short-term, repetitive acidic stimulation can enhance a cell’s capacity to withstand subsequent adverse stress. While AP has demonstrated protective effects in various ischemia-reperfusion (I/R) injury models, its application in kidney injury remains largely unexplored. Methods An AP model was established in human kidney (HK2) cells by treating them with an acidic medium for 12 h, followed by a recovery period with a normal medium for 6 h. To induce hypoxia/reoxygenation (H/R) injury, HK2 cells were subjected to hypoxia for 24 h and reoxygenation for 1 h. In vivo, a mouse model of IPC was established by clamping the bilateral renal pedicles for 15 min, followed by reperfusion for 4 days. Conversely, the I/R model involved clamping the bilateral renal pedicles for 35 min and reperfusion for 24 h. Western blotting was employed to evaluate the expression levels of cleaved caspase 3, cleaved caspase 9, NHE1, KIM1, FAK, and NOX4. A pH-sensitive fluorescent probe was used to measure pH i , while a Hemin/CNF microelectrode monitored kidney tissue pH. Immunofluorescence staining was performed to visualize the localization of NHE1, NOX4, and FAK, along with the actin cytoskeleton structure in HK2 cells. Cell adhesion and scratch assays were conducted to assess cell motility. Results Our findings demonstrated that AP could effectively mitigate H/R injury in HK2 cells. This protective effect and the maintenance of pH i homeostasis by AP involved the upregulation of Na + /H + exchanger 1 (NHE1) expression and activity. The activity of NHE1 was regulated by dynamic changes in pH i -dependent phosphorylation of Focal Adhesion Kinase (FAK) at Y397. This process was associated with NOX4-mediated reactive oxygen species (ROS) production. Furthermore, AP induced the co-localization of FAK, NOX4, and NHE1 in focal adhesions, promoting cytoskeletal remodeling and enhancing cell adhesion and migration capabilities. Conclusions This study provides compelling evidence that AP maintains pH i homeostasis and promotes cytoskeletal remodeling through FAK/NOX4/NHE1 signaling. This signaling pathway ultimately contributes to alleviated H/R injury in HK2 cells.

Ischemia-reperfusion injury (IRI) is the main cause of acute kidney injury (AKI), and there is no effective therapy. Microenvironmental acidification is generally observed in ischemic tissues. Acid-sensing ion channel 1a (ASIC1a) can be activated by a decrease in extracellular pH which mediates neuronal IRI. Our previous study demonstrated that, ASIC1a inhibition alleviates renal IRI. However, the underlying mechanisms have not been fully elucidated. In this study, we determined that renal tubule-specific deletion of ASIC1a in mice (ASIC1afl/fl/CDH16cre) attenuated renal IRI, and reduced the expression of NLRP3, ASC, cleaved-caspase-1, GSDMD-N, and IL-1β. Consistent with these in vivo results, inhibition of ASIC1a by the specific inhibitor PcTx-1 protected HK-2 cells from hypoxia/reoxygenation (H/R) injury, and suppressed H/R-induced NLRP3 inflammasome activation. Mechanistically, the activation of ASIC1a by either IRI or H/R induced the phosphorylation of NF-κB p65, which translocates to the nucleus and promotes the transcription of NLRP3 and pro-IL-1β. Blocking NF-κB by treatment with BAY 11-7082 validated the roles of H/R and acidosis in NLRP3 inflammasome activation. This further confirmed that ASIC1a promotes NLRP3 inflammasome activation, which requires the NF-κB pathway. In conclusion, our study suggests that ASIC1a contributes to renal IRI by affecting the NF-κB/NLRP3 inflammasome pathway. Therefore, ASIC1a may be a potential therapeutic target for AKI.Key messagesKnockout of ASIC1a attenuated renal ischemia-reperfusion injury.ASIC1a promoted the NF-κB pathway and NLRP3 inflammasome activation.Inhibition of the NF-κB mitigated the NLRP3 inflammasome activation induced by ASIC1a.

EEG signals exhibit commonality and variability across subjects, sessions, and tasks. But most existing EEG studies focus on mean group effects (commonality) by averaging signals over trials and subjects. The substantial intra- and inter-subject variability of EEG have often been overlooked. The recently significant technological advances in machine learning, especially deep learning, have brought technological innovations to EEG signal application in many aspects, but there are still great challenges in cross-session, cross-task, and cross-subject EEG decoding. In this work, an EEG-based biometric competition based on a large-scale M CV (A Multi-subject, Multi-session, and Multi-task Database for investigation of EEG Commonality and Variability) database was launched to better characterize and harness the intra- and inter-subject variability and promote the development of machine learning algorithm in this field. In the M CV database, EEG signals were recorded from 106 subjects, of which 95 subjects repeated two sessions of the experiments on different days. The whole experiment consisted of 6 paradigms, including resting-state, transient-state sensory, steady-state sensory, cognitive oddball, motor execution, and steady-state sensory with selective attention with 14 types of EEG signals, 120000 epochs. Two learning tasks (identification and verification), performance metrics, and baseline methods were introduced in the competition. In general, the proposed M CV dataset and the EEG-based biometric competition aim to provide the opportunity to develop advanced machine learning algorithms for achieving an in-depth understanding of the commonality and variability of EEG signals across subjects, sessions, and tasks.