A Dual Robust Graph Neural Network Against Graph Adversarial Attacks.

Graph Neural Networks (GNNs) have gained widespread usage and achieved remarkable success in various real-world applications. Nevertheless, recent studies reveal the vulnerability of GNNs to graph adversarial attacks that fool them by modifying graph structure. This vulnerability undermines the robustness of GNNs and poses significant security and privacy risks across various applications. Hence, it is crucial to develop robust GNN models that can effectively defend against such attacks. One simple approach is to remodel the graph. However, most existing methods cannot fully preserve the similarity relationship among the original nodes while learning the node representation required for reweighting the edges. Furthermore, they lack supervision information regarding adversarial perturbations, hampering their ability to recognize adversarial edges. To address these limitations, we propose a novel Dual Robust Graph Neural Network (DualRGNN) against graph adversarial attacks. DualRGNN first incorporates a node-similarity-preserving graph refining (SPGR) module to prune and refine the graph based on the learned node representations, which contain the original nodes' similarity relationships, weakening the poisoning of graph adversarial attacks on graph data. DualRGNN then employs an adversarial-supervised graph attention (ASGAT) network to enhance the model's capability in identifying adversarial edges by treating these edges as supervised signals. Through extensive experiments conducted on four benchmark datasets, DualRGNN has demonstrated remarkable robustness against various graph adversarial attacks.

A survey on few-shot class-incremental learning.

Large deep learning models are impressive, but they struggle when real-time data is not available. Few-shot class-incremental learning (FSCIL) poses a significant challenge for deep neural networks to learn new tasks from just a few labeled samples without forgetting the previously learned ones. This setup can easily leads to catastrophic forgetting and overfitting problems, severely affecting model performance. Studying FSCIL helps overcome deep learning model limitations on data volume and acquisition time, while improving practicality and adaptability of machine learning models. This paper provides a comprehensive survey on FSCIL. Unlike previous surveys, we aim to synthesize few-shot learning and incremental learning, focusing on introducing FSCIL from two perspectives, while reviewing over 30 theoretical research studies and more than 20 applied research studies. From the theoretical perspective, we provide a novel categorization approach that divides the field into five subcategories, including traditional machine learning methods, meta learning-based methods, feature and feature space-based methods, replay-based methods, and dynamic network structure-based methods. We also evaluate the performance of recent theoretical research on benchmark datasets of FSCIL. From the application perspective, FSCIL has achieved impressive achievements in various fields of computer vision such as image classification, object detection, and image segmentation, as well as in natural language processing and graph. We summarize the important applications. Finally, we point out potential future research directions, including applications, problem setups, and theory development. Overall, this paper offers a comprehensive analysis of the latest advances in FSCIL from a methodological, performance, and application perspective.

Balance guided incomplete multi-view spectral clustering.

There is a large volume of incomplete multi-view data in the real-world. How to partition these incomplete multi-view data is an urgent realistic problem since almost all of the conventional multi-view clustering methods are inapplicable to cases with missing views. In this paper, a novel graph learning-based incomplete multi-view clustering (IMVC) method is proposed to address this issue. Different from existing works, our method aims at learning a common consensus graph from all incomplete views and obtaining a clustering indicator matrix in a unified framework. To achieve a stable clustering result, a relaxed spectral clustering model is introduced to obtain a probability consensus representation with all positive elements that reflect the data clustering result. Considering the different contributions of views to the clustering task, a weighted multi-view learning mechanism is introduced to automatically balance the effects of different views in model optimization. In this way, the intrinsic information of the incomplete multi-view data can be fully exploited. The experiments on several incomplete multi-view datasets show that our method outperforms the compared state-of-the-art clustering methods, which demonstrates the effectiveness of our method for IMVC.

Incomplete multi-view clustering network via nonlinear manifold embedding and probability-induced loss.

Incomplete multi-view clustering, which included missing data in different views, is more challenging than multi-view clustering. For the purpose of eliminating the negative influence of incomplete data, researchers have proposed a series of solutions. However, the present incomplete multi-view clustering methods still confront three major issues: (1) The interference of redundant features hinders these methods to learn the most discriminative features. (2) The importance role of local structure is not considered during clustering. (3) These methods fail to utilize data distribution information to guide models update to decrease the effects of outliers and noise. To address above issues, a novel deep clustering network which exerted on incomplete multi-view data was proposed in this paper. We combine multi-view autoencoders with nonlinear manifold embedding method UMAP to extract latent consistent features of incomplete multi-view data. In the clustering method, we introduce Gaussian Mixture Model (GMM) to fit the complex distribution of data and deal with the interference of outliers. In addition, we reasonably utilize the probability distribution information generated by GMM, using probability-induced loss function to integrate feature learning and clustering as a joint framework. In experiments conducted on multiple benchmark datasets, our method captures incomplete multi-view data features effectively and perform excellent.

Two-Layered Microfluidic Devices for High-Throughput Dynamic Analysis of Synthetic Gene Circuits in E. coli.

Escherichia coli is a common chassis for synthetic gene circuit studies. In addition to the dose-response of synthetic gene circuits, the analysis of dynamic responses is also an important part of the future design of more complicated synthetic systems. Recently, microfluidic-based methods have been widely used for the analysis of gene expression dynamics. Here, we established a two-layered microfluidic platform for the systematic characterization of synthetic gene circuits (eight strains in eight different culture environments could be observed simultaneously with a 5 min time resolution). With this platform, both dose responses and dynamic responses with a high temporal resolution could be easily derived for further analysis. A controlled environment ensures the stability of the bacterial growth rate, excluding changes in gene expression dynamics caused by changes of the growth dilution rate. The precise environmental switch and automatic micrograph shooting ensured that there was nearly no time lag between the inducer addition and the data recording. We studied four four-node incoherent-feedforward-loop (IFFL) networks with different operators using this device. The experimental results showed that as the effect of inhibition increased, two of the IFFL networks generated pulselike dynamic gene expressions in the range of the inducer concentrations, which was different from the dynamics of the two other circuits with only a simple pattern of rising to the platform. Through fitting the dose-response curves and the dynamic response curves, corresponding parameters were derived and introduced to a simple model that could qualitatively explain the generation of pulse dynamics.

Perioperative administration of methylprednisolone was associated with postoperative pulmonary complications in elderly patients undergoing hip fracture surgery.

BACKGROUND: Postoperative pulmonary complications (PPCs) seriously affect the postoperative prognosis of elderly patients underwent hip fracture surgery. Although methylprednisolone is increasingly used, the association between perioperative methylprednisolone and PPCs is still controversial. The study aims to determine whether perioperative administration of methylprednisolone is associated with PPCs in elderly patients during hip fracture surgery. PATIENTS AND METHODS: In this retrospective cohort study, records of 584 patients (≥ 65 years) who underwent hip fracture surgery between January 2013 and October 2020 were extracted. Univariate and multivariate regression analysis were performed to identify the risk factors for PPCs. To further explore the association between administration of methylprednisolone and PPCs, 53 patients received methylprednisolone and 53 patients without methylprednisolone were matched for the confounding factors using propensity score matching (PSM) analysis. The odds ratios (OR) and 95% confidence intervals (CI) for the above variables were analyzed. RESULTS: The incidence of PPCs during postoperative hospitalization was 6.83% (38/556) among the elderly patients following hip fracture surgery. Patients with PPCs had higher postoperative mortality rate, longer hospital stay, more hospitalization cost, and higher incidence of cardiac arrest (all P < 0.05). Multivariate logistic regression analysis showed that age, hypertension, hypoglycemia, hypoproteinemia and perioperative methylprednisolone were independent risk factors for PPCs. Moreover, administration of methylprednisolone was significantly correlated with PPCs both before PSM adjustment (OR = 3.25; 95% CI, 1.67 to 6.33; P = 0.001) and after PSM adjustment (OR = 6.68; 95% CI, 1.40 to 31.82; P = 0.017). CONCLUSION: Perioperative administration of methylprednisolone is a risk factor for PPCs in elderly patients undergoing hip fracture surgery.

Discriminant Geometrical and Statistical Alignment With Density Peaks for Domain Adaptation.

Unsupervised domain adaptation (DA) aims to perform classification tasks on the target domain by leveraging rich labeled data in the existing source domain. The key insight of DA is to reduce domain divergence by learning domain-invariant features or transferable instances. Despite its rapid development, there still exist several challenges to explore. At the feature level, aligning both domains only in a single way (i.e., geometrical or statistical) has limited ability to reduce the domain divergence. At the instance level, interfering instances often obstruct learning a discriminant subspace when performing the geometrical alignment. At the classifier level, only minimizing the empirical risk on the source domain may result in a negative transfer. To tackle these challenges, this article proposes a novel DA method, called discriminant geometrical and statistical alignment (DGSA). DGSA first aligns the geometrical structure of both domains by projecting original space into a Grassmann manifold, then matches the statistical distributions of both domains by minimizing their maximum mean discrepancy on the manifold. In the former step, DGSA only selects the density peaks to learn the Grassmann manifold and so to reduce the influences of interfering instances. In addition, DGSA exploits the high-confidence soft labels of target landmarks to learn a more discriminant manifold. In the latter step, a structural risk minimization (SRM) classifier is learned to match the distributions (both marginal and conditional) and predict the target labels at the same time. Extensive experiments on objection recognition and human activity recognition tasks demonstrate that DGSA can achieve better performance than the comparison methods.

Dual Alignment for Partial Domain Adaptation.

Partial domain adaptation (PDA) aims to transfer knowledge from a label-rich source domain to a label-scarce target domain based on an assumption that the source label space subsumes the target label space. The major challenge is to promote positive transfer in the shared label space and circumvent negative transfer caused by the large mismatch across different label spaces. In this article, we propose a dual alignment approach for PDA (DAPDA), including three components: 1) a feature extractor extracts source and target features by the Siamese network; 2) a reweighting network produces "hard" labels, class-level weights for source features and "soft" labels, instance-level weights for target features; 3) a dual alignment network aligns intra domain and interdomain distributions. Specifically, the intra domain alignment aims to minimize the intraclass variances to enhance the intraclass compactness in both domains, and interdomain alignment attempts to reduce the discrepancies across domains by domain-wise and class-wise adaptations. The negative transfer can be alleviated by down-weighting source features with nonshared labels. The positive transfer can be enhanced by upweighting source features with shared labels. The adaptation can be achieved by minimizing the discrepancies based on class-weighted source data with hard labels and instance-weighed target data with soft labels. The effectiveness of our method has been demonstrated by outperforming state-of-the-art PDA methods on several benchmark datasets.

Highly ordered mesoporous carbon nanomatrix as a new approach to improve the oral absorption of the water-insoluble drug, simvastatin.

Three different kinds of highly ordered mesoporous carbon (HMC) matrices with different morphologies (hexagonal, spherical and fibrous), particle sizes (700 nm, 400-900 nm and 1-4 µm) and pore diameters were compared as drug carriers for a model drug, simvastatin (SIM). The physicochemical properties of the SIM-loaded composites were studied using field emission scanning electron microscopy (FESEM), specific surface area analysis, differential scanning calorimetry (DSC), wide-angle X-ray scattering (WAXS), HPLC, solubility measurement and dissolution testing. Furthermore, the oral bioavailability of SIM-loaded SHMC (spherical HMC nanomatrix) in beagle dogs was compared with that of the reference formulation (Zocor®). The results obtained showed that SIM molecules are encapsulated in a noncrystalline state due to geometric confinement in the nanopores of HMC. In vitro dissolution testing showed that the dissolution rate of SIM released from monodispersed SHMC was significantly faster compared with that of crystalline SIM and other SIM-loaded composites. In addition, in vivo bioavailability study demonstrated that the relative bioavailability of SIM and SIM ß-hydroxy acid (an active metabolite of SIM) for SIM-loaded SHMC formulation was 138.42% and 163.55%, respectively. In conclusion, monodispersed SHMC appear to be a more promising candidate as a new oral drug delivery vehicle providing a rapid drug release and enhanced oral bioavailability.

A novel SOD1 mutation in amyotrophic lateral sclerosis with a distinct clinical phenotype.

Familial amyotrophic lateral sclerosis (FALS) accounts for about 5% of cases of the neurodegenerative disorder ALS. At least 100 Cu/Zn superoxide dismutase (SOD1) genetic mutations have been associated with FALS. We identified a FALS family in China with an atypical clinical phenotype. To investigate the SOD1 gene mutations in this family, five exons of the SOD1 gene from each living patient were amplified by PCR and screened by SSCP and direct DNA sequencing. SSCP analysis demonstrated a mutation in exon 2 of SOD1, and DNA sequencing demonstrated the presence of an insertion mutation in exon 2 that has not been reported previously. The mutant SOD1 gene encodes a truncated protein of 35 amino acid residues compared to the normal SOD1 protein of 153 amino acids. In conclusion, The SOD1 exon 2 mutation is likely to be the etiological factor of ALS in this family.

The 9L(LUC)/Wistar rat glioma model is not suitable for immunotherapy.

The availability of a well-characterized animal brain tumor model will play an important role in identifying treatments for human brain tumors. Wistar rats bearing 9L glioma cells can develop solid, well-circumcised tumors, and may be a useful animal model for the evaluation of various therapeutic approaches for gliosarcomas. In this study, the 9L/Wistar rat glioma model was produced by intracerebral implantation of 9L(LUC) glioma cells syngenic to Fischer 344 (F344) rats. Bioluminescence imaging showed that tumors progressively grew from day 7 to day 21 in 9L(LUC)/F344 rats, and tumor regression was found in some 9L(LUC)/Wistar rats. Hematoxylin-eosin staining verified that intracranial tumors were gliomas. Immunohistochemistry results demonstrated that no CD4- and CD8-positive cells were found in the syngeneic 9L(LUC)/F344 model. However, many infiltrating CD4- and CD8-positive cells were observed within the tumors of the 9L(LUC)/Wistar model. Our data suggests that compared with 9L/F344 rats, 9L glioma Wistar rats may not be suitable for evaluating brain glioma immunotherapies, even though the model induced an immune response and exhibited tumor regression.

Enhanced angiogenesis and astrocyte activation by ecdysterone treatment in a focal cerebral ischemia rat model.

BACKGROUND AND PURPOSE: We reported previously that ecdysterone (EDS) improves neurologic function after experimental stroke. However, the underlying mechanism remained unclear. The present study was conducted to test whether ecdysterone improves neurologic function by enhancing astrocyte activation and angiogenesis after focal cerebral ischemia in rats. METHODS: Focal cerebral ischemia model was conducted by middle cerebral artery occlusion (MCAO). EDS was intraperitoneally injected at 20 mg kg1 daily for 7 days after MCAO. Neurologic recovery was assessed using the neurologic severity scores. Microvessel density and GFAP expression were detected with immunostaining and analyzed quantitatively with image system. RESULTS: Treatment with EDS significantly improved functional recovery, along with increases in density of cerebral microvessels and astrocyte activation. Microvessel density was significantly higher in EDS treated group than in ischemia control group at all time points, and reached a peak on day 14. EDS treated group had substantial increment in GFAP immunoreactive cells, darker staining color, more and longer nerve processes, higher GFAP expression and area of immunoreactive cells at each time point. CONCLUSION: Our data suggest that EDS treatment enhanced angiogenesis and astrocyte activation which could contribute to functional recovery.

Development of transgenic endothelial progenitor cell-seeded stents.

Endothelial progenitor cell (EPC)-seeded intravascular stents may reduce or prevent in-stent restenosis. A20 can play an important role for preventing vascular restenosis. Therefore, it is very important how to enhance the seeding efficiency of A20-modified EPCs on the stent for preventing in-stent restenosis. To approach this problem, we developed a novel transgenic EPC-seeded stent and evaluated its feasibility and efficiency. EPCs were isolated and purified from umbilical blood using immunomagnetic beads and then transfected with the A20 gene. One stent type (type 1) was coated with EDC cross-linked collagen, and another stent type (type 2) was coated with EDC cross-linked collagen and bound to the CD34 antibody using the bifunctional coupling agent N-succinmidyl3-(2-pyridyldithio) propionate (SPDP). Then, the stents were seeded with EPCs transfected with the A20 gene. The stents were implanted in biological artificial vessels, and cell adhesion was determined in a flow chamber. Cell growth was also measured. EPCs were transfected successfully with the A20 gene. The cells covered both types of stents with favorable biological function. After placement in a flow chamber, the number of cells attached to type 1 stents significantly dropped and their distribution was scattered. Type 2 stents were basically covered with cells and there were more cells on type 2 stents than on type 1 stents (p < 0.01). Collagen-coupled antibody effectively improves the seeding of transgenic EPCs, offering a new choice of stents to prevent restenosis caused by vascular disease after interventional treatment.

[Study on arousal effect of Orexin-A in rat in coma due to ischemic brain injury].

OBJECTIVE: To reproduce an ischemic brain injury coma model and explore the arousal effect of Orexin-A. METHODS: An ischemic brain injury coma model was reproduced in rats by partial four-vessel occlusion (4-VO with a needle of 0.60 mm in diameter in the lumen to create stenosis of the internal carotid arteries). One hundred and twenty minutes after the onset of coma, Orexin-A or its antagonist (SB-334867) was given intraventricularly, and the time of coma and changes in electroencephalogram (ECG) were observed, and the unit discharge of neurons in the prefrontal cortex was recorded. RESULTS: Partial occlusion of four internal carotid arteries, reducing the lumens to 0.60 mm, could prolong the time of coma to 6-8 hours with the rats still alive. The duration of coma showed a significant difference compared with that in rats who underwent 0.45 mm or 0.70 mm stenosis of the internal carotid arteries (F=344.43, P<0.01). Intraventricular Orexin-A in a dose of 4 nmol/10 microl could obviously decrease the duration of coma with a decrease in alpha wave and increase in unit discharge rate of neurons in coma rats (P<0.05 or P<0.01), but no significant change was observed when the dose was 2 nmol/10 microl. CONCLUSION: (1)Creating stenosis of all four internal carotid arteries is suitable to reproduce ischemic brain injury with coma in rats. (2)Intracerebroventricular injection of Orexin-A (4 nmol) has a potent arousal effect on ischemic brain injury coma in rats.

[Identification of the mutation of SOD1 gene in a familial amyotrophic lateral sclerosis].

OBJECTIVE: To identify the mutation of Cu/Zn superoxide dismutase(SOD1) gene in an amyotrophic lateral sclerosis (ALS) family with unique phenotype. METHODS: Five exons of SOD1 gene were amplified by PCR. The differences of these products were analyzed by PCR-single strand conformation polymorphism and visualized by silver staining. RESULTS: Abnormal bands were found in exons 2 and 5 of SOD1 gene in several familial members. DNA sequence analysis verified that a base pair insertion occurred in the codon area of exon 2 and in the intron area of exon 5. And the insertion mutation of exon 2 led to a frameshift mutation and premature stop. It is a new type of SOD1 mutation which may be associated with familial amyotrophic lateral sclerosis. CONCLUSION: Insertion mutation of exon 2 may be responsible for the disease of an ALS family in Chongqing.