Hierarchical Augmentation and Distillation for Class Incremental Audio-Visual Video Recognition.

Audio-visual video recognition (AVVR) aims to integrate audio and visual clues to categorize videos accurately. While existing methods train AVVR models using provided datasets and achieve satisfactory results, they struggle to retain historical class knowledge when confronted with new classes in real-world situations. Currently, there are no dedicated methods for addressing this problem, so this paper concentrates on exploring Class Incremental Audio-Visual Video Recognition (CIAVVR). For CIAVVR, since both stored data and learned model of past classes contain historical knowledge, the core challenge is how to capture past data knowledge and past model knowledge to prevent catastrophic forgetting. As audio-visual data and model inherently contain hierarchical structures, i.e., model embodies low-level and high-level semantic information, and data comprises snippet-level, video-level, and distribution-level spatial information, it is essential to fully exploit the hierarchical data structure for data knowledge preservation and hierarchical model structure for model knowledge preservation. However, current image class incremental learning methods do not explicitly consider these hierarchical structures in model and data. Consequently, we introduce Hierarchical Augmentation and Distillation (HAD), which comprises the Hierarchical Augmentation Module (HAM) and Hierarchical Distillation Module (HDM) to efficiently utilize the hierarchical structure of data and models, respectively. Specifically, HAM implements a novel augmentation strategy, segmental feature augmentation, to preserve hierarchical model knowledge. Meanwhile, HDM introduces newly designed hierarchical (video-distribution) logical distillation and hierarchical (snippet-video) correlative distillation to capture and maintain the hierarchical intra-sample knowledge of each data and the hierarchical inter-sample knowledge between data, respectively. Evaluations on four benchmarks (AVE, AVK-100, AVK-200, and AVK-400) demonstrate that the proposed HAD effectively captures hierarchical information in both data and models, resulting in better preservation of historical class knowledge and improved performance. Furthermore, we provide a theoretical analysis to support the necessity of the segmental feature augmentation strategy.

A Platform for Site-Specific DNA-Antibody Bioconjugation by Using Benzoylacrylic-Labelled Oligonucleotides.

Many bioconjugation strategies for DNA oligonucleotides and antibodies suffer limitations, such as site-specificity, stoichiometry and hydrolytic instability of the conjugates, which makes them unsuitable for biological applications. Here, we report a new platform for the preparation of DNA-antibody bioconjugates with a simple benzoylacrylic acid pentafluorophenyl ester reagent. Benzoylacrylic-labelled oligonucleotides prepared with this reagent can be site-specifically conjugated to a range of proteins and antibodies through accessible cysteine residues. The homogeneity of the prepared DNA-antibody bioconjugates was confirmed by a new LC-MS protocol and the bioconjugate probes were used in fluorescence or super-resolution microscopy cell imaging experiments. This work demonstrates the versatility and robustness of our bioconjugation protocol that gives site-specific, well-defined and plasma-stable DNA-antibody bioconjugates for biological applications.

Attention-Based Multi-Source Domain Adaptation.

Multi-source domain adaptation (MSDA) aims to transfer knowledge from multi-source domains to one target domain. Inspired by single-source domain adaptation, existing methods solve MSDA by aligning the data distributions between the target domain and each source domain. However, aligning the target domain with the dissimilar source domain would harm the representation learning. To address the above issue, an intuitive motivation of MSDA is using the attention mechanism to enhance the positive effects of the similar domains, and suppress the negative effects of the dissimilar domains. Therefore, we propose Attention-Based Multi-Source Domain Adaptation (ABMSDA) by considering the domain correlations to alleviate the effects caused by dissimilar domains. To obtain the domain correlations between source and target domains, ABMSDA firstly trains a domain recognition model to calculate the probability that the target images belong to each source domain. Based on the domain correlations, Weighted Moment Distance (WMD) is proposed to pay more attention on the source domains with higher similarities. Furthermore, Attentive Classification Loss (ACL) is developed to constrain that the feature extractor can generate the alignment and discriminative visual representations. The evaluations on two benchmarks demonstrate the effectiveness of the proposed model, e.g., an average of 6.1% improvement on the challenging DomainNet dataset.

A General in Vitro Assay for Studying Enzymatic Activities of the Ubiquitin System.

The ubiquitin (Ub) system regulates a wide range of cellular signaling pathways. Several hundred E1, E2, and E3 enzymes are together responsible for protein ubiquitination, thereby controlling cellular activities. Due to the numerous enzymes and processes involved, studies of ubiquitination activities have been challenging. We here report a novel Förster resonance energy transfer (FRET)-based assay for studying the in vitro kinetics of ubiquitination. FRET is established upon binding of fluorophore-labeled Ub to eGFP-tagged ZnUBP, a domain that exclusively binds unconjugated Ub. We name this assay the free Ub sensor system (FUSS). Using Uba1, UbcH5, and CHIP as model E1, E2, and E3 enzymes, respectively, we demonstrate that ubiquitination results in decreasing FRET efficiency, from which reaction rates can be determined. Further treatment with USP21, a deubiquitinase, leads to increased FRET efficiency, confirming the reversibility of the assay. We subsequently use this assay to show that increasing the concentration of CHIP or UbcH5 but not Uba1 enhances ubiquitination rates and develop a novel machine learning approach to model ubiquitination. The overall ubiquitination activity is also increased upon incubation with tau, a substrate of CHIP. Our data together demonstrate the versatile applications of a novel ubiquitination assay that does not require labeling of E1, E2, E3, or substrates and is thus likely compatible with any E1-E2-E3 combinations.

Tumour necrosis factor induces increased production of extracellular amyloid-ß- and α-synuclein-containing aggregates by human Alzheimer's disease neurons.

In addition to increased aberrant protein aggregation, inflammation has been proposed as a key element in the pathogenesis and progression of Alzheimer's disease. How inflammation interacts with other disease pathways and how protein aggregation increases during disease are not clear. We used single-molecule imaging approaches and membrane permeabilization assays to determine the effect of chronic exposure to tumour necrosis factor, a master proinflammatory cytokine, on protein aggregation in human-induced pluripotent stem cell-derived neurons harbouring monogenic Alzheimer's disease mutations. We report that exposure of Alzheimer's disease neurons, but not control neurons, to tumour necrosis factor induces substantial production of extracellular protein aggregates. Aggregates from Alzheimer's disease neurons are composed of amyloid-ß and α-synuclein and induce significant permeabilization of lipid membranes in an assay of pathogenicity. These findings provide support for a causal relationship between two crucial processes in Alzheimer's disease pathogenesis and suggest that targeting inflammation, particularly tumour necrosis factor, may have beneficial downstream effects on ameliorating aberrant protein aggregation and accumulation.

The Efficacy of Cognitive Training for Elderly Chinese Individuals with Mild Cognitive Impairment.

The age of the population is shifting toward the elderly range, which may lead to an increased risk of mild cognitive impairment (MCI). The aims of this study are to evaluate the cognitive function in elderly people using the Montreal Cognitive Assessment (MoCA), to identify the relationship between cognitive function and different characteristics, and to evaluate the efficacy of the intervention after six months of cognitive training. In this study, we included 2886 subjects aged ≧60 years in the baseline survey, and 140 subjects with MCI who participated in the baseline survey were randomly divided into an intervention group (N = 70) and a control group (N = 70). The control group was not provided any intervention measures, and the intervention group was administered cognitive training. The education level, monthly income, sleep time, exercise time, reading times, and time spent engaging in community activities and performing housework were positively correlated with MoCA scores, but age was negatively correlated with MoCA scores. The total MoCA score of the intervention group increased from 19.77 ± 2.24 points to 21.09 ± 2.20 points after six months of cognitive training, but the score of the control group decreased from 20.41 ± 2.10 points to 19.17 ± 2.57 points. The two-way repeated-measures ANOVA revealed a very significant effect of the interaction between time and cognitive training on the total MoCA score. Seventeen participants in the intervention group improved to normal levels, and no participants progressed to dementia after six months of cognitive training. Thus, the efficacy of the intervention was statistically significant. Our study concludes that older age is associated with a cognitive decline. Factors that are more likely to protect against cognitive decline included a higher education level and monthly income, sufficient sleep time, regular physical exercise and reading, frequently engaging in community activities, and continuing to perform housework. Moreover, the cognitive training intervention is effective and may help to decrease the deterioration of cognitive function in patients with MCI, and the interaction between intervention time and cognitive training significantly improves cognitive function.

Relationship Between the Incidence of Road Traffic Accidents, Psychological Characteristics, and Genotype in Bus Drivers in a Chinese Population.

BACKGROUND The aim of this study was to determine the association between the incidence of road traffic accidents, psychological characteristics, and genotype in bus drivers in a Chinese population. MATERIAL AND METHODS Bus drivers who had been involved in road traffic accidents (n=106) (the study group), and bus drivers with no history of road traffic accidents (n=106) (the control group) completed demographic questionnaires, the Eysenck Personality Questionnaire (EPQ) and the Type-A behavior pattern (TABP) evaluation. Serum 5-hydroxytryptamine (5-HT) (serotonin), and 5-hydroxytryptophan (5-HTP) levels were measured by high-performance liquid chromatography-fluorescent detection (HPLC-FLD). Serotonin transporter promoter-linked polymorphism region (5-HTTLPR) and the 521 C/T single nucleotide polymorphism (SNP) in the regulatory region of the dopamine D4 receptor gene (DRD4-521 C/T) were measured using polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP). RESULTS After accounting for potential confounders, extroversion, psychopathy, neuroticism and time hurrying (impatience) were significant factors associated with road traffic accidents in bus drivers (adjusted OR: 1.268, 95% CI: 1.133-1.419; adjusted OR: 1.177, 95% CI: 1.028-1.347; adjusted OR: 1.092, 95% CI: 1.005-1.187; adjusted OR: 1.123, 95% CI: 1.025-1.230, respectively). Reduced serum levels of 5-HT and 5-HTP were significantly associated with the incidence of road traffic accidents (adjusted OR: 0.985, 95% CI: 0.973-0.997; adjusted OR: 0.982, 95% CI: 0.969-0.994, respectively). CONCLUSIONS Psychological characteristics associated with the 5-HTTLPR and DRD4-521 C/T genotypes, including extroversion, psychopathy, neuroticism, and time hurrying (impatience), and low serum levels of 5-HT and 5-HTP in bus drivers were associated with an increased risk of road traffic accidents.

Nanoscopic Characterisation of Individual Endogenous Protein Aggregates in Human Neuronal Cells.

The aberrant misfolding and subsequent conversion of monomeric protein into amyloid aggregates characterises many neurodegenerative disorders, including Parkinson's and Alzheimer's diseases. These aggregates are highly heterogeneous in structure, generally of low abundance and typically smaller than the diffraction limit of light (≈250 nm). To overcome the challenges these characteristics pose to the study of endogenous aggregates formed in cells, we have developed a method to characterise them at the nanometre scale without the need for a conjugated fluorophore. Using a combination of DNA PAINT and an amyloid-specific aptamer, we demonstrate that this technique is able to detect and super-resolve a range of aggregated species, including those formed by α-synuclein and amyloid-ß. Additionally, this method enables endogenous protein aggregates within cells to be characterised. We found that neuronal cells derived from patients with Parkinson's disease contain a larger number of protein aggregates than those from healthy controls.