Isobavachin attenuates osteoclastogenesis and periodontitis-induced bone loss by inhibiting cellular iron accumulation and mitochondrial biogenesis.

As bone-resorbing cells rich in mitochondria, osteoclasts require high iron uptake to promote mitochondrial biogenesis and maintain a high-energy metabolic state for active bone resorption. Given that abnormal osteoclast formation and activation leads to imbalanced bone remodeling and osteolytic bone loss, osteoclasts may be crucial targets for treating osteolytic diseases such as periodontitis. Isobavachin (IBA), a natural flavonoid compound, has been confirmed to be an inhibitor of receptor activator of nuclear factor κB ligand (RANKL)-induced osteoclast differentiation from bone marrow-derived macrophages (BMMs). However, its effects on periodontitis-induced bone loss and the potential mechanism of its anti-osteoclastogenesis effect remain unclear. Our study demonstrated that IBA suppressed RANKL-induced osteoclastogenesis in BMMs and RAW264.7 cells and inhibited osteoclast-mediated bone resorption in vitro. Transcriptomic analysis indicated that iron homeostasis and reactive oxygen species (ROS) metabolic process were enriched among the differentially expressed genes following IBA treatment. IBA exerted its anti-osteoclastogenesis effect by inhibiting iron accumulation in osteoclasts. Mechanistically, IBA attenuated iron accumulation in RANKL-induced osteoclasts by inhibiting the mitogen-activated protein kinase (MAPK) pathway to upregulate ferroportin1 (Fpn1) expression and promote Fpn1-mediated intracellular iron efflux. We also found that IBA inhibited mitochondrial biogenesis and function, and reduced RANKL-induced ROS generation in osteoclasts. Furthermore, IBA attenuated periodontitis-induced bone loss by reducing osteoclastogenesis in vivo. Overall, these results suggest that IBA may serve as a promising therapeutic strategy for bone diseases characterized by osteoclastic bone resorption.

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.

Source-Guided Target Feature Reconstruction for Cross-Domain Classification and Detection.

Existing cross-domain classification and detection methods usually apply a consistency constraint between the target sample and its self-augmentation for unsupervised learning without considering the essential source knowledge. In this paper, we propose a Source-guided Target Feature Reconstruction (STFR) module for cross-domain visual tasks, which applies source visual words to reconstruct the target features. Since the reconstructed target features contain the source knowledge, they can be treated as a bridge to connect the source and target domains. Therefore, using them for consistency learning can enhance the target representation and reduce the domain bias. Technically, source visual words are selected and updated according to the source feature distribution, and applied to reconstruct the given target feature via a weighted combination strategy. After that, consistency constraints are built between the reconstructed and original target features for domain alignment. Furthermore, STFR is connected with the optimal transportation algorithm theoretically, which explains the rationality of the proposed module. Extensive experiments on nine benchmarks and two cross-domain visual tasks prove the effectiveness of the proposed STFR module, e.g., 1) cross-domain image classification: obtaining average accuracy of 91.0%, 73.9%, and 87.4% on Office-31, Office-Home, and VisDA-2017, respectively; 2) cross-domain object detection: obtaining mAP of 44.50% on Cityscapes → Foggy Cityscapes, AP on car of 78.10% on Cityscapes → KITTI, MR -2 of 8.63%, 12.27%, 22.10%, and 40.58% on COCOPersons → Caltech, CityPersons → Caltech, COCOPersons → CityPersons, and Caltech → CityPersons, respectively.

IRX5 suppresses osteogenic differentiation of hBMSCs by inhibiting protein synthesis.

In our previous study, IRX5 has been revealed a significant role in adipogenesis of hBMSCs. Considering the expansion of adipose tissue in bone marrow in aged and ovariectomy-related osteoporosis, the effect of IRX5 on the osteogenesis of BMSCs still needs to be elucidated. In vivo, models of aging-induced and ovariectomy-induced osteoporotic mice, and in vitro studies of IRX5 gene gain- and loss-of-function in hBMSCs were employed. Histology, immunofluorescence, qRT-PCR, and Western blot analysis were performed to detect the functions of IRX5 in hBMSCs osteogenic differentiation. RNA-seq, transmission electron microscopy, Seahorse mito-stress assay, and Surface Sensing of Translation assay were conducted to explore the effect of mammalian/mechanistic target of rapamycin (mTOR)-mediated ribosomal translation and mitochondrial functions in the regulation of hBMSCs differentiation by IRX5. As a result, elevated IRX5 protein expression levels were observed in the bone marrow of osteoporotic mice compared to normal mice. IRX5 overexpression attenuated osteogenic processes, whereas IRX5 knockdown resulted in enhanced osteogenesis in hBMSCs. RNA-seq and enrichment analysis unveiled that IRX5 overexpression exerted inhibitory effects on ribosomal translation and mitochondrial functions. Furthermore, the application of the mTOR activator, MHY1485, effectively reversed the inhibitory impact of IRX5 on osteogenesis and mitochondrial functions in hBMSCs. In summary, our findings suggest that IRX5 restricts mTOR-mediated ribosomal translation, consequently impairing mitochondrial OxPhos, which in turn results in osteogenic dysfunction of hBMSCs.

Sulforaphene Inhibits Periodontitis through Regulating Macrophage Polarization via Upregulating Dendritic Cell Immunoreceptor.

Periodontitis is one of the most prevalent chronic inflammatory diseases that may eventually lead to the loss of teeth. Macrophage polarization plays an important role in the development of periodontitis, and several naturally occurring food compounds have recently been reported to regulate macrophage polarization. In this study, we aimed to investigate the therapeutic potential of sulforaphene (SFE) in macrophage polarization and its impact on periodontitis. Through in vitro and in vivo experiments, our study demonstrated that SFE effectively inhibits M1 polarization while promoting M2 polarization, ultimately leading to the suppression of periodontitis. Transcriptome sequencing showed that SFE significantly upregulated the expression of dendritic cell immunoreceptor (DCIR, also known as CLEC4A2). We further validated the crucial role of DCIR in macrophage polarization through knockdown and overexpression experiments and demonstrated that SFE regulates macrophage polarization by upregulating DCIR expression. In summary, the results of this study suggest that SFE can regulate macrophage polarization and inhibit periodontitis. Moreover, this research identified DCIR (dendritic cell immunoreceptor) as a potential novel target for regulating macrophage polarization. These findings provide new insights into the treatment of periodontitis and other immune-related diseases.

Sulforaphene suppresses RANKL-induced osteoclastogenesis and LPS-induced bone erosion by activating Nrf2 signaling pathway.

BACKGROUND AND PURPOSE: Inflammatory disorders have been found to induce bone loss through sustained and persistent activation of osteoclast differentiation, leading to heightened bone resorption. The current pharmacological interventions for combating bone loss to harbor adverse effects or contraindications. There is a pressing need to identify drugs with fewer side effects. EXPERIMENTAL APPROACH: The effect and underlying mechanism of sulforaphene (LFS) on osteoclast differentiation were illustrated in vitro and in vivo with RANKL-induced Raw264.7 cell line osteoclastogenesis and lipopolysaccharide (LPS)-induced bone erosion model. KEY RESULTS: In this study, LFS has been shown to effectively impede the formation of mature osteoclasts induced from both Raw264.7 cell line and bone marrow macrophages (BMMs), mainly at the early stage. Further mechanistic investigations uncovered that LFS suppressed AKT phosphorylation. SC-79, a potent AKT activator, was found to reverse the inhibitory impact of LFS on osteoclast differentiation. Moreover, transcriptome sequencing analysis revealed that treatment with LFS led to a significant upregulation in the expression of nuclear factor erythroid 2-related factor 2 (Nrf2) and antioxidant-related genes. Then it's validated that LFS could promote NRF2 expression and nuclear translocation, as well as effectively resist oxidative stress. NRF2 knockdown reversed the suppression effect of LFS on osteoclast differentiation. In vivo experiments provide convincing evidence that LFS is protective against LPS-induced inflammatory osteolysis. CONCLUSION AND IMPLICATIONS: These well-grounded and promising findings suggest LFS as a promising agent to addressing oxidative-stress related diseases and bone loss disorders.

Anti-Porphyromonas gingivalis nanotherapy for maintaining bacterial homeostasis in periodontitis.

Periodontitis is caused by oral flora imbalance, which leads to immune imbalance. Porphyromonas gingivalis is a keystone pathogen in periodontitis, causing the blooming of inflammophilic microbes, and becoming dormant to resist antibiotics. Targeted interventions are needed to destroy this pathogen and collapse its inflammophilic flora. Therefore, a targeting nanoagent antibody-conjugated liposomal drug carrier with ginsenoside Rh2 (A-L-R) was developed for pleiotropic benefits. A-L-R showed high quality in high-performance liquid chromatography (HPLC), Fourier transform infrared (FTIR), and transmission electron microscope (TEM) detection. Only P. gingivalis was influenced by A-L-R, as shown by live/dead cell staining and a series of antimicrobial effects assays. With fluorescence in situ hybridization (FISH) staining and in propidium monoazide-quantitative polymerase chain reaction (PMA-qPCR), the clearance of P. gingivalis by A-L-R was more than for other groups, and only the proportion of P. gingivalis was reduced by A-L-R in monospecies culture. Moreover, in a periodontitis model, A-L-R targeted P. gingivalis with high efficiency and low toxicity, maintaining homeostasis with a relatively stable oral microflora. This targeting nanomedicine offers new strategies for periodontitis therapy, providing a foundation for the prevention and treatment of periodontitis.

Heat Shock Protein Inhibitors Show Synergistic Antibacterial Effects with Photodynamic Therapy on Caries-Related Streptococci In Vitro and In Vivo.

Caries are chronic infections in which the cariogenic biofilm plays a critical role in disease occurrence and progression. Photodynamic therapy (PDT) is a new effective treatment that is receiving wide attention in the antibacterial field, but it can lead to the upregulation of heat shock proteins (HSPs), which enhances bacterial resistance. Herein, we incorporated HSP inhibitors with PDT to evaluate the effect on Streptococcus mutans, Streptococcus sobrinus, and Streptococcus sanguinis under planktonic conditions and on cariogenic biofilms. Additionally, a model of caries was established in 2-week-old rats, and anticaries properties were evaluated by Keyes' scoring. Importantly, the combination of HSP inhibitors and PDT had outstanding efficiency in inhibiting the growth of tested Streptococcus strains and the formation of either monomicrobial or multispecies biofilms in vitro. In addition, the quantity of colonized streptococci and the severity of carious lesions were also distinctly suppressed in vivo. Overall, the synergistic application of HSP inhibitors and PDT has promising potential in the prevention and treatment of dental caries. IMPORTANCE Effective therapies for the prevention and control of caries are urgently needed. Cariogenic streptococci play a key role in the occurrence and progression of caries. Recently, photodynamic therapy has been demonstrated to have good antibacterial efficiency, but it can cause a heat shock response in bacteria, which may weaken its practical effects. We indicate here an effective therapeutic strategy of combining heat shock protein inhibitors and photodynamic therapy, which shows excellent inhibition toward three dominant streptococci related to caries and suppression of carious progression in a rat model. Further development for clinical application is promising.

Dual Instance-Consistent Network for Cross-Domain Object Detection.

Cross-domain object detection aims to transfer knowledge from a labeled dataset to an unlabeled dataset. Most existing methods apply a unified embedding model to generate the tightly coupled source and target descriptions for domain alignment, leading to the destroyed feature distribution of the target domain because the embedding model is mainly controlled by the source domain. To reduce the representation bias of the target domain, we apply two independent networks to extract two types of discriminative descriptions with mutual consistency, i.e., a novel Dual Instance-Consistent Network (DICN) is proposed for cross-domain object detection. Especially, Dual Instance-Consistent Module containing the instance mutual consistency between Primary Network and Auxiliary Network is applied to align two domains, where Primary and Auxiliary Networks are used to obtain the source-specific and target-specific information, respectively. The instance mutual consistency consists of two terms: feature consistency and detection consistency, which is applied to align the instance feature and the output of detection head, respectively. With the instance mutual consistency, optimizing the Primary (Auxiliary) Network only with source (target) images by fixing the Auxiliary (Primary) Network can generate the source(target)-specific description. Extensive experiments on several benchmarks demonstrate the effectiveness of the proposed DICN, e.g., obtaining mAP of 44.10% for Cityscapes → Foggy Cityscapes, AP on car of 76.50% for Cityscapes → KITTI, MR -2 of 8.87%, 12.66%, 22.27%, and 42.06% for COCOPersons → Caltech, CityPersons → Caltech, COCOPersons → CityPersons, and Caltech → CityPersons, respectively.

Profiling the oral microbiomes in patients with Alzheimer's disease.

OBJECTIVES: To analyse the characteristics of the oral microbiomes and expected to find biomarkers about Alzheimer's disease (AD). SUBJECTS AND METHODS: AD patients (n = 26) and cognitive intact people (n = 26) were examined for cognition, depression, oral health and collected saliva and gingival crevicular fluid (GCF) in the morning. Full-length 16S rRNA gene was amplified and sequencing was performed using the PacBio platform. RESULTS: The predominant bacterium of salivary microbiome and periodontal microbiome from AD patients was Streptococcus oralis and Porphyromonas gingivalis, respectively. With respect to ß diversity analysis, there was a significance difference in periodontal microbiome between AD patients and cognitively intact subjects. The relative abundance of Veillonella parvula significantly increased in oral microbiomes from AD patients. Interestingly, the dominant species were different between early-onset AD and late-onset AD patients. Moreover, the predominant species were changed as the clinical severity of AD. Furthermore, the correlation analysis revealed that V. parvula was associated with AD in both saliva and GCF and that P. gingivalis was associated with AD only in GCF. CONCLUSIONS: In this study, the microbiome community of oral microbes was altered in AD patients and periodontal microbiome was sensitive to cognition changes. Moreover, V. parvula and P. gingivalis were associated with AD.

Ginsenoside Rb3 inhibits osteoclastogenesis via ERK/NF-κB signaling pathway in vitro and in vivo.

OBJECTIVE: The objective of the study was to determine the anti-osteoclastogenic potential of ginsenoside Rb3 for the treatment of periodontitis. METHODS: The anti-osteoclastogenic effect was determined using RANKL-induced RAW264.7 cells and murine bone marrow-derived macrophages followed by TRAP and phalloidin staining. Expression of osteoclastogenesis-related genes and proteins were examined by qPCR and WB. Activation of signaling pathways was detected by WB and IHC techniques. Experimental periodontitis rat model was built up by gingival injections of P. gingivalis LPS. After 21 days of Rb3 treatment, rats were sacrificed for micro-CT, IHC, H&E, and TRAP staining analyses. RESULTS: Rb3 dramatically inhibits RANKL-induced osteoclastogenesis. Nfatc1, Mmp9, Ctsk, Acp5 mRNA, and MMP9, CTSK proteins were dose-dependently downregulated by Rb3 pretreatment. WB results revealed that Rb3 suppressed activations of p38 MAPK, ERK, and p65 NF-κB, and the inhibition of ERK was most pronounced. Consistently, IHC analysis revealed that p-ERK was highly expressed in alveolar bone surface, blood vessels, odontoblasts, and gingival epithelia, which were notably suppressed by Rb3 treatment. H&E staining and micro-CT analyses showed that Rb3 significantly attenuated gingivitis and alveolar bone resorption in rats. CONCLUSION: Rb3 inhibits RANKL-induced osteoclastogenesis and attenuates P. gingivalis LPS-induced gingivitis and alveolar bone resorption in rats via ERK/NF-κB signaling pathway.

The dynamic communities of oral microbiome in neonates.

The oral microbiome, associated with both oral disease and systemic disease, is in dynamic status along the whole life, and many factors including maternal microbiomes could impact the oral microbiome. While fewer studies have been conducted to study the characteristics of the oral microbiome in neonates and the associated maternal factors. Hence, we collected the microbiome of 15 mother-infant pairs across multiple body sites from birth up to 4 days postpartum and used high-throughput sequencing to characterize the microbiomes in mothers and their neonates. The oral microbiome in the neonates changed obviously during the 4 days after birth. Many bacteria originating from the vagina, skin, and environment disappeared in oral cavity over time, such as Prevotella bivia and Prevotella jejuni. Meanwhile, Staphylococcus epidermidis RP62A phage SP-beta, predominate bacterium in maternal skin microbiome and Streptococcus unclassified, main bacterium in vaginal microbiome, obviously increased in neonatal oral microbiome as time went on. Interestingly, as time progressed, the composition of the oral microbiome in the neonates was more similar to that of the milk microbiome in their mothers. Moreover, we found that the changes in the predominant bacteria in the neonates were in line with those in the neonates exposed to the environment. Together, these data described the sharp dynamics of the oral microbiome in neonates and the importance of maternal efforts in the development of the neonatal microbiome.

IRX5 promotes adipogenesis of hMSCs by repressing glycolysis.

Iroquois homeobox transcription factor 5 (IRX5) plays a pivotal role in extramedullary adipogenesis, but little is known about the effects of IRX5 on adipogenesis of human bone marrow-derived mesenchymal stem cells (hMSCs). In this study, we aimed to determine the effect of IRX5 on hMSCs adipogenesis. By means of qPCR analysis, we determined that IRX5 expression was elevated during adipogenic commitment of hMSCs. The biologic role of IRX5 was further investigated by employing a gain/loss-of-function strategy using an in vitro lentivirus-based system. IRX5 overexpression promoted adipogenesis whereas IRX5 knockdown reduced the adipogenic phenotype. RNA-seq and metabolomics revealed that IRX5 overexpression repressed glycolysis. Dual-luciferase assay results showed that IRX5 overexpression transcriptionally activates peroxisome proliferator-activated receptor gamma coactivator (PGC-1α). Metformin and PGC-1α inhibitor reversed IRX5-induced adipogenesis and glycolytic inhibition. Collectively, IRX5 facilitates adipogenic differentiation of hMSCs by transcriptionally regulating PGC-1α and inhibiting glycolysis, revealing a potential target to control bone marrow-derived mesenchymal stem cells (BMSCs) fate decision and bone homeostasis.

Application of Ginsenoside Rd in Periodontitis With Inhibitory Effects on Pathogenicity, Inflammation, and Bone Resorption.

Periodontitis is a worldwide oral disease induced by the interaction of subgingival bacteria and host response and is characterized by local inflammation, bone resorption, and tooth loss. Ginsenoside Rd (Rd) is a biologically active component derived from Panax ginseng and has been demonstrated to exert antibacterial and anti-inflammatory activities. This study aims to investigate the inhibitory efficiency of Rd towards Porphyromonas gingivalis (P. gingivalis), periodontal inflammatory response, and osteoclastogenesis in vitro and to further validate the results in a mouse periodontitis model, thus, evaluate the potential effects of Rd on the control and prevention of periodontitis. According to the results, Rd exerted excellent antibacterial activities against planktonic P. gingivalis, along with attenuating P. gingivalis virulence and inhibiting its biofilms. Meanwhile, the inflammatory cytokine production and osteoclastogenesis were remarkably inhibited by Rd both in vitro and in vivo. Furthermore, Rd efficiently ameliorated the subgingival P. gingivalis abundance and suppressed the alveolar bone resorption in vivo as well. In conclusion, Rd has the potential to be developed as a promising medication in the control and prevention of periodontitis.

Bioresponsive nanotherapy for preventing dental caries by inhibiting multispecies cariogenic biofilms.

Early childhood caries (ECC) is a public healthcare concern that greatly reduces the quality of life of young children. As a leading factor of ECC, cariogenic biofilms are composed of acidogenic/aciduric pathogens and extracellular polysaccharides (EPSs), creating an acidic and protected microenvironment. Antimicrobial photodynamic therapy (aPDT) is a noninvasive, painless, and efficient therapeutic approach that is suitable for treating ECC. However, due to the hyperfine structure of cariogenic biofilms, most photosensitizers (PSs) could not access and penetrate deeply in biofilms, which dramatically hamper their efficiency in the clinic. Herein, bioresponsive nanoparticle loaded with chlorin e6 (MPP-Ce6) is developed, which largely increases the penetration depth (by over 75%) and retention (by over 100%) of PS in the biofilm compared with free Ce6. Furthermore, MPP-Ce6-mediated aPDT not only kills the bacteria in preformed biofilms but also inhibits multispecies biofilm formation. A rampant caries model is established to mimic ECC in vivo, where the population of cariogenic bacteria is decreased to 10% after MPP-Ce6-mediated aPDT. Importantly, the number and severity of carious lesions are efficiently reduced via Keyes' scoring and micro-CT analysis. This simple but effective strategy can serve as a promising approach for daily oral hygiene in preventing ECC.

Lung cancer scRNA-seq and lipidomics reveal aberrant lipid metabolism for early-stage diagnosis.

Lung cancer is the leading cause of cancer mortality, and early detection is key to improving survival. However, there are no reliable blood-based tests currently available for early-stage lung cancer diagnosis. Here, we performed single-cell RNA sequencing of different early-stage lung cancers and found that lipid metabolism was broadly dysregulated in different cell types, with glycerophospholipid metabolism as the most altered lipid metabolism-related pathway. Untargeted lipidomics was carried out in an exploratory cohort of 311 participants. Through support vector machine algorithm-based and mass spectrum-based feature selection, we identified nine lipids (lysophosphatidylcholines 16:0, 18:0, and 20:4; phosphatidylcholines 16:0-18:1, 16:0-18:2, 18:0-18:1, 18:0-18:2, and 16:0-22:6; and triglycerides 16:0-18:1-18:1) as the features most important for early-stage cancer detection. Using these nine features, we developed a liquid chromatography-mass spectrometry (MS)-based targeted assay using multiple reaction monitoring. This target assay achieved 100.00% specificity on an independent validation cohort. In a hospital-based lung cancer screening cohort of 1036 participants examined by low-dose computed tomography and a prospective clinical cohort containing 109 participants, the assay reached more than 90.00% sensitivity and 92.00% specificity. Accordingly, matrix-assisted laser desorption/ionization MS imaging confirmed that the selected lipids were differentially expressed in early-stage lung cancer tissues in situ. This method, designated as Lung Cancer Artificial Intelligence Detector, may be useful for early detection of lung cancer or large-scale screening of high-risk populations for cancer prevention.

Metabolic detection and systems analyses of pancreatic ductal adenocarcinoma through machine learning, lipidomics, and multi-omics.

Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal cancers, characterized by rapid progression, metastasis, and difficulty in diagnosis. However, there are no effective liquid-based testing methods available for PDAC detection. Here we introduce a minimally invasive approach that uses machine learning (ML) and lipidomics to detect PDAC. Through greedy algorithm and mass spectrum feature selection, we optimized 17 characteristic metabolites as detection features and developed a liquid chromatography-mass spectrometry-based targeted assay. In this study, 1033 patients with PDAC at various stages were examined. This approach has achieved 86.74% accuracy with an area under curve (AUC) of 0.9351 in the large external validation cohort and 85.00% accuracy with 0.9389 AUC in the prospective clinical cohort. Accordingly, single-cell sequencing, proteomics, and mass spectrometry imaging were applied and revealed notable alterations of selected lipids in PDAC tissues. We propose that the ML-aided lipidomics approach be used for early detection of PDAC.

TEST: Triplet Ensemble Student-Teacher Model for Unsupervised Person Re-Identification.

The self-ensembling methods have achieved amazing performance for semi-supervised representation learning and domain adaptation. However, the disadvantage of these methods is that the teacher network is tightly coupled with the student network, which limits the descriptive ability of the self-ensembling model. To overcome the coupling effect between the teacher network and the student network, we propose a novel Triplet Ensemble Student-Teacher (TEST) model for unsupervised person re-identification, which consists of one teacher network T and two student networks S1 and S2 . Similar to the traditional self-ensembling model, the student network S1 is applied to update the teacher network T . Furthermore, a closed-loop learning mechanism is built in the TEST model by imposing an ensemble consistent constraint between T and S2 , and performing a heterogeneous co-teaching procedure between S1 and S2 . With the closed-loop learning mechanism, the TEST model can loosen the constraint between the teacher T and the student S1 , and enhance the descriptive ability of S1 . Besides, the knowledge exchange between S1 and S2 can ensure that the two student networks can elegantly deal with the noisy labels and avoid coupling. By training the TEST model with the clustering-generated pseudo labels, we can achieve effective and robust representation learning for unsupervised person re-identification. The evaluations on three widely-used benchmarks show that our approach can achieve significant performance compared with state-of-the-art methods.

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.

SAN: Selective Alignment Network for Cross-Domain Pedestrian Detection.

Cross-domain pedestrian detection, which has been attracting much attention, assumes that the training and test images are drawn from different data distributions. Existing methods focus on aligning the descriptions of whole candidate instances between source and target domains. Since there exists a giant visual difference among the candidate instances, aligning whole candidate instances between two domains cannot overcome the inter-instance difference. Compared with aligning the whole candidate instances, we consider that aligning each type of instances separately is a more reasonable manner. Therefore, we propose a novel Selective Alignment Network for cross-domain pedestrian detection, which consists of three components: a Base Detector, an Image-Level Adaptation Network, and an Instance-Level Adaptation Network. The Image-Level Adaptation Network and Instance-Level Adaptation Network can be regarded as the global-level and local-level alignments, respectively. Similar to the Faster R-CNN, the Base Detector, which is composed of a Feature module, an RPN module and a Detection module, is used to infer a robust pedestrian detector with the annotated source data. Once obtaining the image description extracted by the Feature module, the Image-Level Adaptation Network is proposed to align the image description with an adversarial domain classifier. Given the candidate proposals generated by the RPN module, the Instance-Level Adaptation Network firstly clusters the source candidate proposals into several groups according to their visual features, and thus generates the pseudo label for each candidate proposal. After generating the pseudo labels, we align the source and target domains by maximizing and minimizing the discrepancy between the prediction of two classifiers iteratively. Extensive evaluations on several benchmarks demonstrate the effectiveness of the proposed approach for cross-domain pedestrian detection.