MARK4 controls ischaemic heart failure through microtubule detyrosination.

Myocardial infarction is a major cause of premature death in adults. Compromised cardiac function after myocardial infarction leads to chronic heart failure with systemic health complications and a high mortality rate1. Effective therapeutic strategies are needed to improve the recovery of cardiac function after myocardial infarction. More specifically, there is a major unmet need for a new class of drugs that can improve cardiomyocyte contractility, because inotropic therapies that are currently available have been associated with high morbidity and mortality in patients with systolic heart failure2,3 or have shown a very modest reduction of risk of heart failure4. Microtubule detyrosination is emerging as an important mechanism for the regulation of cardiomyocyte contractility5. Here we show that deficiency of microtubule-affinity regulating kinase 4 (MARK4) substantially limits the reduction in the left ventricular ejection fraction after acute myocardial infarction in mice, without affecting infarct size or cardiac remodelling. Mechanistically, we provide evidence that MARK4 regulates cardiomyocyte contractility by promoting phosphorylation of microtubule-associated protein 4 (MAP4), which facilitates the access of vasohibin 2 (VASH2)-a tubulin carboxypeptidase-to microtubules for the detyrosination of α-tubulin. Our results show how the detyrosination of microtubules in cardiomyocytes is finely tuned by MARK4 to regulate cardiac inotropy, and identify MARK4 as a promising therapeutic target for improving cardiac function after myocardial infarction.

Characterization of renal artery variation in patients with clear cell renal cell carcinoma and the predictive value of accessory renal artery in pathological grading of renal cell carcinoma: a retrospective and observational study.

OBJECTIVE: To explore the characteristics of renal artery variation in patients with renal cell carcinoma and to evaluate the predicting value of accessory renal artery in the pathological grading of renal cell carcinoma. METHODS: The clinicopathological data of patients with clear cell renal cell carcinoma diagnosed in the Department of Urology of the First Hospital of Shanxi Medical University from September 2019 to March 2023 were retrospectively analyzed. All patients underwent visual three-dimensional model reconstruction from computed tomography images. All kidneys were divided into two groups: the affected kidney and the healthy kidney, and the incidence of renal artery variation in the two groups was analyzed. Then, according to the existence of accessory renal artery in the affected kidney, the patients were divided into two groups, and the relationship between accessory renal artery and clinicopathological features of patients with clear cell renal cell carcinoma was analyzed. Finally, univariate and multivariate logistic regression analyses were performed to determine the predictors of Fuhrman grading of clear cell renal cell carcinoma, and the predictive ability of the model was evaluated by the receiver operating characteristic curve. RESULTS: The incidence of renal artery variation and accessory renal artery in the affected kidney was significantly higher than them in the healthy kidney. The patients with accessory renal artery in the affected kidney had larger tumor maximum diameter, higher Fuhrman grade and more exophytic growth. The presence of accessory renal artery on the affected kidney and the maximum diameter of tumor are independent predictors of high-grade renal cell carcinoma. The receiver operating characteristic curve suggests that the model has a good predictive ability. CONCLUSION: The existence of accessory renal artery on the affected kidney may be related to the occurrence and development of clear cell renal cell carcinoma, and can better predict Fuhrman grade of clear cell renal cell carcinoma. The finding provides a reference for the future diagnostic evaluation of RCC, and provides a new direction for the study of the pathogenesis of RCC.

Robust nontarget DNA-triggered catalytic hairpin assembly amplification strategy for the improved sensing of microRNA in complex biological matrices.

A simple but robust fluorescence strategy based on a nontarget DNA-triggered catalytic hairpin assembly (CHA) was constructed to probe microRNA-21 (miR-21). A short ssDNA rather than degradable target miRNA was employed as an initiator. Two molecular beacons needed to assist the CHA process were simplified to avoid unfavorable nonspecific interactions. In the presence of the target, the initiator was released from a partially duplex and triggered the cyclic CHA reaction, resulting in a significantly amplified optical readout. A wide linear range from 0.1 pM to 1000 pM for the sensing of miR-21 in buffer was achieved with a low detection limit of 0.76 pM. Fortunately, this strategy demonstrated an obviously improved performance for miR-21 detection in diluted serum. The fluorescence signals were enhanced remarkably and the sensitivity was further improved to 0.12 pM in 10% serum. The stability for miR-21 quantification and the capability for the analysis of single nucleotide polymorphisms (SNPs) were also improved greatly. More importantly, the biosensor could be applied to image miR-21 in different living tumor cells with high resolution, illustrating its promising potential for the assay of miRNAs in various complex situations for early-stage disease diagnosis and biological studies in cells.

Preoperative serum low-density lipoprotein cholesterol is an independent prognostic factor in patients with renal cell carcinoma after nephrectomy.

BACKGROUND: Little is known about the association between the preoperative low-density lipoprotein cholesterol (LDL-C) level and prognosis in patients with renal cell carcinoma (RCC) after nephrectomy, and its prognostic value needs to be elucidated. METHODS: The clinical and follow-up data of 737 RCC patients who underwent nephrectomy were retrospectively analyzed. The optimal cut-off LDL-C level was determined using X-tile, and then patients were divided into low and high LDL-C groups. The association between LDL-C levels and survival of RCC patients was assessed using the Kaplan-Meier method and Cox regression analysis. RESULTS: The optimal cut-off LDL-C level was 1.93 mmol/L, and patients were divided into the low (≤ 1.93 mmol/L) and high LDL-C (> 1.93 mmol/L) groups. The Kaplan-Meier analysis showed that patients in the low LDL-C group had significantly shorter overall survival (OS), cancer-specific survival (CSS) and recurrence-free survival (RFS) than those in the high LDL-C group (P = 0.001, P = 0.001, and P = 0.003, respectively). The COX univariate analysis showed that the preoperative LDL-C level was closely associated with OS, CSS, and RFS in RCC patients (P = 0.002, P = 0.003, and P = 0.005, respectively). The multivariate analysis showed that the preoperative LDL-C level was an independent factor for predicting survival (OS, CSS and RFS) in RCC patients after nephrectomy. The low preoperative LDL-C levels predicted worse OS (hazard ratio [HR]: 2.337; 95% confidence interval [CI]: 1.192-4.581; P = 0.013), CSS (HR: 3.347; 95% CI: 1.515-7.392; P = 0.003), and RFS (HR: 2.207; 95% CI: 1.178-4.132; P = 0.013). CONCLUSIONS: The preoperative LDL-C level is an independent factor for the prognosis of RCC patients after nephrectomy, and low preoperative LDL-C levels predict worse survival (OS, CSS, and RFS).

Cepharanthine Ameliorates Pulmonary Fibrosis by Inhibiting the NF-κB/NLRP3 Pathway, Fibroblast-to-Myofibroblast Transition and Inflammation.

Pulmonary fibrosis (PF) is one of the sequelae of Corona Virus Disease 2019 (COVID-19), and currently, lung transplantation is the only viable treatment option. Hence, other effective treatments are urgently required. We investigated the therapeutic effects of an approved botanical drug, cepharanthine (CEP), in a cell culture model of transforming growth factor-ß1 (TGF-ß1) and bleomycin (BLM)-induced pulmonary fibrosis rat models both in vitro and in vivo. In this study, CEP and pirfenidone (PFD) suppressed BLM-induced lung tissue inflammation, proliferation of blue collagen fibers, and damage to lung structures in vivo. Furthermore, we also found increased collagen deposition marked by α-smooth muscle actin (α-SMA) and Collagen Type I Alpha 1 (COL1A1), which was significantly alleviated by the addition of PFD and CEP. Moreover, we elucidated the underlying mechanism of CEP against PF in vitro. Various assays confirmed that CEP reduced the viability and migration and promoted apoptosis of myofibroblasts. The expression levels of myofibroblast markers, including COL1A1, vimentin, α-SMA, and Matrix Metallopeptidase 2 (MMP2), were also suppressed by CEP. Simultaneously, CEP significantly suppressed the elevated Phospho-NF-κB p65 (p-p65)/NF-κB p65 (p65) ratio, NOD-like receptor thermal protein domain associated protein 3 (NLRP3) levels, and elevated inhibitor of NF-κB Alpha (IκBα) degradation and reversed the progression of PF. Hence, our study demonstrated that CEP prevented myofibroblast activation and treated BLM-induced pulmonary fibrosis in a dose-dependent manner by regulating nuclear factor kappa-B (NF-κB)/ NLRP3 signaling, thereby suggesting that CEP has potential clinical application in pulmonary fibrosis in the future.

GLH: From Global to Local Gradient Attacks with High-Frequency Momentum Guidance for Object Detection.

The adversarial attack is crucial to improving the robustness of deep learning models; they help improve the interpretability of deep learning and also increase the security of the models in real-world applications. However, existing attack algorithms mainly focus on image classification tasks, and they lack research targeting object detection. Adversarial attacks against image classification are global-based with no focus on the intrinsic features of the image. In other words, they generate perturbations that cover the whole image, and each added perturbation is quantitative and undifferentiated. In contrast, we propose a global-to-local adversarial attack based on object detection, which destroys important perceptual features of the object. More specifically, we differentially extract gradient features as a proportion of perturbation additions to generate adversarial samples, as the magnitude of the gradient is highly correlated with the model's point of interest. In addition, we reduce unnecessary perturbations by dynamically suppressing excessive perturbations to generate high-quality adversarial samples. After that, we improve the effectiveness of the attack using the high-frequency feature gradient as a motivation to guide the next gradient attack. Numerous experiments and evaluations have demonstrated the effectiveness and superior performance of our from global to Local gradient attacks with high-frequency momentum guidance (GLH), which is more effective than previous attacks. Our generated adversarial samples also have excellent black-box attack ability.

Tralesinidase Alfa Enzyme Replacement Therapy Prevents Disease Manifestations in a Canine Model of Mucopolysaccharidosis Type IIIB.

Mucopolysaccharidosis type IIIB (MPS IIIB; Sanfilippo syndrome B; OMIM #252920) is a lethal, pediatric, neuropathic, autosomal recessive, and lysosomal storage disease with no approved therapy. Patients are deficient in the activity of N-acetyl-alpha-glucosaminidase (NAGLU; EC 3.2.150), necessary for normal lysosomal degradation of the glycosaminoglycan heparan sulfate (HS). Tralesinidase alfa (TA), a fusion protein comprised of recombinant human NAGLU and a modified human insulin-like growth factor 2, is in development as an enzyme replacement therapy that is administered via intracerebroventricular (ICV) infusion, thus circumventing the blood brain barrier. Previous studies have confirmed ICV infusion results in widespread distribution of TA throughout the brains of mice and nonhuman primates. We assessed the long-term tolerability, pharmacology, and clinical efficacy of TA in a canine model of MPS IIIB over a 20-month study. Long-term administration of TA was well tolerated as compared with administration of vehicle. TA was widely distributed across brain regions, which was confirmed in a follow-up 8-week pharmacokinetic/pharmacodynamic study. MPS IIIB dogs treated for up to 20 months had near-normal levels of HS and nonreducing ends of HS in cerebrospinal fluid and central nervous system (CNS) tissues. TA-treated MPS IIIB dogs performed better on cognitive tests and had improved CNS pathology and decreased cerebellar volume loss relative to vehicle-treated MPS IIIB dogs. These findings demonstrate the ability of TA to prevent or limit the biochemical, pathologic, and cognitive manifestations of canine MPS IIIB disease, thus providing support of its potential long-term tolerability and efficacy in MPS IIIB subjects. SIGNIFICANCE STATEMENT: This work illustrates the efficacy and tolerability of tralesinidase alfa as a potential therapeutic for patients with mucopolysaccharidosis type IIIB (MPS IIIB) by documenting that administration to the central nervous system of MPS IIIB dogs prevents the accumulation of disease-associated glycosaminoglycans in lysosomes, hepatomegaly, cerebellar atrophy, and cognitive decline.

Underwater image enhancement using adaptive color restoration and dehazing.

Underwater images captured by optical cameras can be degraded by light attenuation and scattering, which leads to deteriorated visual image quality. The technique of underwater image enhancement plays an important role in a wide range of subsequent applications such as image segmentation and object detection. To address this issue, we propose an underwater image enhancement framework which consists of an adaptive color restoration module and a haze-line based dehazing module. First, we employ an adaptive color restoration method to compensate the deteriorated color channels and restore the colors. The color restoration module consists of three steps: background light estimation, color recognition, and color compensation. The background light estimation determines the image is blueish or greenish, and the compensation is applied in red-green or red-blue channels. Second, the haze-line technique is employed to remove the haze and enhance the image details. Experimental results show that the proposed method can restore the color and remove the haze at the same time, and it also outperforms several state-of-the-art methods on three publicly available datasets. Moreover, experiments on an underwater object detection dataset show that the proposed underwater image enhancement method is able to improve the accuracy of the subsequent underwater object detection framework.

SecDH: Security of COVID-19 images based on data hiding with PCA.

Nowadays, image security and copyright protection become challenging, especially after the COVID-19 pandemic. In the paper, we develop SecDH as a medical data hiding scheme, which can guarantee the security and copyright protection of the COVID-19 images. Firstly, the cover image is normalized, which offers high resistance against the geometric attacks. Secondly, the normalized principal component as embedding factor is computed, which are calculated based on principal component analysis (PCA) between cover and mark image. Thirdly, the medical image is invisibly marked with secret mark based on normalized component, redundant discrete wavelet transform (RDWT) and randomized singular value decomposition (RSVD) is introduced. Finally, Arnold cat map scheme employed to ensure the security of the watermarking system. Under the experimental evaluation, our SecDH tool is not only imperceptible, but also has a satisfactory advantage in robustness and security compared with the traditional watermarking schemes.

Achieving efficacy in subjects with sustained pegvaliase-neutralizing antibody responses.

Pegvaliase (Palynziq®) is an enzyme substitution therapy using PEGylated recombinant Anabaena variabilis phenylalanine ammonia lyase (PAL) to reduce blood phenylalanine (Phe) levels in adults with phenylketonuria (PKU). In Phase 3 clinical studies, all subjects treated with pegvaliase developed anti-drug antibodies. To specifically evaluate pegvaliase-neutralizing antibodies (NAbs) and assess impact on pegvaliase efficacy, a novel hybrid ligand-binding/tandem mass spectrometry NAb assay was developed. Analysis of Phase 3 study samples revealed that pegvaliase NAb titers developed during early treatment (≤6 months after treatment initiation), and then plateaued and persisted in the majority of subjects during late treatment (>6 months). Subjects with the lowest/undetectable NAb titers had relatively high plasma pegvaliase concentrations and experienced the most rapid decline in blood Phe concentrations at relatively low pegvaliase dose concentrations. In contrast, subjects with higher NAb titers generally had lower plasma pegvaliase concentrations on similar low doses, with little change in blood Phe concentrations. However, with additional time on treatment and individualized dose titration, the majority of subjects achieved substantial and sustained blood Phe reduction, including those with higher NAb titers. Moreover, after maturation of the anti-pegvaliase immune response, NAb titers were stable over time and did not rise in response to dose increases; thus, subjects did not require additional dose increases to maintain reduction in blood Phe.

Three-Dimensional Reconstruction with a Laser Line Based on Image In-Painting and Multi-Spectral Photometric Stereo.

This paper presents a multi-spectral photometric stereo (MPS) method based on image in-painting, which can reconstruct the shape using a multi-spectral image with a laser line. One of the difficulties in multi-spectral photometric stereo is to extract the laser line because the required illumination for MPS, e.g., red, green, and blue light, may pollute the laser color. Unlike previous methods, through the improvement of the network proposed by Isola, a Generative Adversarial Network based on image in-painting was proposed, to separate a multi-spectral image with a laser line into a clean laser image and an uncorrupted multi-spectral image without the laser line. Then these results were substituted into the method proposed by Fan to obtain high-precision 3D reconstruction results. To make the proposed method applicable to real-world objects, a rendered image dataset obtained using the rendering models in ShapeNet has been used for training the network. Evaluation using the rendered images and real-world images shows the superiority of the proposed approach over several previous methods.

Impacts of COVID-19 and anti-pandemic policies on urban transport-an empirical study in China.

The COVID-19 pandemic that began in the last quarter of 2019 seriously impacted the transportation industry. Countries around the world adopted various restrictions and policies to prevent the spread of the pandemic, which resulted in a sharp drop in the demand for transportation. China was the first country to detect the pandemic and the fastest to recover. Existing policies and impacts were reviewed to analyze the impact of the pandemic on China's urban transportation sector and propose measures that may be taken to reduce the impact of COVID-19. This study reviews the impact on urban transportation system operations and how government should respond to a viral pandemic. The recovery measures during and after the pandemic and their hierarchical response system are analyzed. Furthermore, to empirically explore the effect of the recovery measures, this study adopted the Event Study Methodology (ESM) to quantitatively analyze the impact of the epidemic as well as anti-pandemic policies on the traffic flow sequence in the resurgence of COVID-19 in Beijing. The research findings provided solid policy implications and experiences for constructing sustainable urban transportation system and improve flexibility, reliability, and resilience of traffic governance in post-pandemic era.

FarpScusn: Fully Anonymous Routing Protocol with Self-Healing Capability in Unstable Sensor Networks.

Anonymous technology is an effective way for protecting users' privacy. Anonymity in sensor networks is to prevent the unauthorized third party from revealing the identities of the communication parties. While, in unstable wireless sensor networks, frequent topology changes often lead to route-failure in anonymous communication. To deal with the problems of anonymous route-failure in unstable sensor networks, in this paper we propose a fully anonymous routing protocol with self-healing capability in unstable sensor networks by constructing a new key agreement scheme and proposing an anonymous identity scheme. The proposed protocol maintains full anonymity of sensor nodes with the self-healing capability of anonymous routes. The results from the performance analysis show that the proposed self-healing anonymity-focused protocol achieves full anonymity of source nodes, destination nodes, and communication association.

C2DAN: An Improved Deep Adaptation Network with Domain Confusion and Classifier Adaptation.

Deep neural networks have been successfully applied in domain adaptation which uses the labeled data of source domain to supplement useful information for target domain. Deep Adaptation Network (DAN) is one of these efficient frameworks, it utilizes Multi-Kernel Maximum Mean Discrepancy (MK-MMD) to align the feature distribution in a reproducing kernel Hilbert space. However, DAN does not perform very well in feature level transfer, and the assumption that source and target domain share classifiers is too strict in different adaptation scenarios. In this paper, we further improve the adaptability of DAN by incorporating Domain Confusion (DC) and Classifier Adaptation (CA). To achieve this, we propose a novel domain adaptation method named C2DAN. Our approach first enables Domain Confusion (DC) by using a domain discriminator for adversarial training. For Classifier Adaptation (CA), a residual block is added to the source domain classifier in order to learn the difference between source classifier and target classifier. Beyond validating our framework on the standard domain adaptation dataset office-31, we also introduce and evaluate on the Comprehensive Cars (CompCars) dataset, and the experiment results demonstrate the effectiveness of the proposed framework C2DAN.

Spatio-Temporal Attention Model for Foreground Detection in Cross-Scene Surveillance Videos.

Foreground detection is an important theme in video surveillance. Conventional background modeling approaches build sophisticated temporal statistical model to detect foreground based on low-level features, while modern semantic/instance segmentation approaches generate high-level foreground annotation, but ignore the temporal relevance among consecutive frames. In this paper, we propose a Spatio-Temporal Attention Model (STAM) for cross-scene foreground detection. To fill the semantic gap between low and high level features, appearance and optical flow features are synthesized by attention modules via the feature learning procedure. Experimental results on CDnet 2014 benchmarks validate it and outperformed many state-of-the-art methods in seven evaluation metrics. With the attention modules and optical flow, its F-measure increased 9 % and 6 % respectively. The model without any tuning showed its cross-scene generalization on Wallflower and PETS datasets. The processing speed was 10.8 fps with the frame size 256 by 256.

Semi-Supervised Generative Adversarial Nets with Multiple Generators for SAR Image Recognition.

As an important model of deep learning, semi-supervised learning models are based on Generative Adversarial Nets (GANs) and have achieved a competitive performance on standard optical images. However, the training of GANs becomes unstable when they are applied to SAR images, which reduces the feature extraction capability of the discriminator in GANs. This paper presents a new semi-supervised GANs with Multiple generators and a classifier (MCGAN). This model improves the stability of training for SAR images by employing multiple generators. A multi-classifier is introduced to the new GANs to utilize the labeled images during the training of the GANs, which shares the low level layers with the discriminator. Then, the layers of the trained discriminator and the classifier construct the recognition network for SAR images after having been finely tuned using a small number of the labeled images. Experiments on the Moving and Stationary Target Acquisition and Recognition (MSTAR) databases show that the proposed recognition network achieves a better and more stable recognition performance than several traditional semi-supervised methods as well as other GANs-based semi-supervised methods.

Adaptive bacteria colony picking in unstructured environments using intensity histogram and unascertained LS-SVM classifier.

Features analysis is an important task which can significantly affect the performance of automatic bacteria colony picking. Unstructured environments also affect the automatic colony screening. This paper presents a novel approach for adaptive colony segmentation in unstructured environments by treating the detected peaks of intensity histograms as a morphological feature of images. In order to avoid disturbing peaks, an entropy based mean shift filter is introduced to smooth images as a preprocessing step. The relevance and importance of these features can be determined in an improved support vector machine classifier using unascertained least square estimation. Experimental results show that the proposed unascertained least square support vector machine (ULSSVM) has better recognition accuracy than the other state-of-the-art techniques, and its training process takes less time than most of the traditional approaches presented in this paper.

Have we found a solution for health misinformation? A ten-year systematic review of health misinformation literature 2013-2022.

BACKGROUND: Health misinformation (HM) has emerged as a prominent social issue in recent years, driven by declining public trust, popularisation of digital media platforms and escalating public health crisis. Since the Covid-19 pandemic, HM has raised critical concerns due to its significant impacts on both individuals and society as a whole. A comprehensive understanding of HM and HM-related studies would be instrumental in identifying possible solutions to address HM and the associated challenges. METHODS: Following the PRISMA procedure, 11,739 papers published from January 2013 to December 2022 were retrieved from five electronic databases, and 813 papers matching the inclusion criteria were retained for further analysis. This article critically reviewed HM-related studies, detailing the factors facilitating HM creation and dissemination, negative impacts of HM, solutions to HM, and research methods employed in those studies. RESULTS: A growing number of studies have focused on HM since 2013. Results of this study highlight that trust plays a significant while latent role in the circuits of HM, facilitating the creation and dissemination of HM, exacerbating the negative impacts of HM and amplifying the difficulty in addressing HM. CONCLUSION: For health authorities and governmental institutions, it is essential to systematically build public trust in order to reduce the probability of individuals acceptation of HM and to improve the effectiveness of misinformation correction. Future studies should pay more attention to the role of trust in how to address HM.

Effects of different contact angles during forefoot running on the stresses of the foot bones: a finite element simulation study.

Introduction: The purpose of this study was to compare the changes in foot at different sole-ground contact angles during forefoot running. This study tried to help forefoot runners better control and improve their technical movements by comparing different sole-ground contact angles. Methods: A male participant of Chinese ethnicity was enlisted for the present study, with a recorded age of 25 years, a height of 183 cm, and a body weight of 80 kg. This study focused on forefoot strike patterns through FE analysis. Results: It can be seen that the peak von Mises stress of M1-5 (Metatarsal) of a (Contact angle: 9.54) is greater than that of b (Contact angle: 7.58) and c (Contact angle: 5.62) in the three cases. On the contrary, the peak von Mises stress of MC (Medial Cuneiform), IC (Intermediate Cuneiform), LC (Lateral Cuneiform), C (Cuboid), N (Navicular), T (Tarsal) in three different cases is opposite, and the peak von Mises stress of c is greater than that of a and b. The peak von Mises stress of b is between a and c. Conclusion: This study found that a reduced sole-ground contact angle may reduce metatarsal stress fractures. Further, a small sole-ground contact angle may not increase ankle joint injury risk during forefoot running. Hence, given the specialized nature of the running shoes designed for forefoot runners, it is plausible that this study may offer novel insights to guide their athletic pursuits.

InA: Inhibition Adaption on pre-trained language models.

Fine-tuning pre-trained language models (LMs) may not always be the most practical approach for downstream tasks. While adaptation fine-tuning methods have shown promising results, a clearer explanation of their mechanisms and further inhibition of the transmission of information is needed. To address this, we propose an Inhibition Adaptation (InA) fine-tuning method that aims to reduce the number of added tunable weights and appropriately reweight knowledge derived from pre-trained LMs. The InA method involves (1) inserting a small trainable vector into each Transformer attention architecture and (2) setting a threshold to directly eliminate irrelevant knowledge. This approach draws inspiration from the shunting inhibition, which allows the inhibition of specific neurons to gate other functional neurons. With the inhibition mechanism, InA achieves competitive or even superior performance compared to other fine-tuning methods on BERT-large, RoBERTa-large, and DeBERTa-large for text classification and question-answering tasks.