EWT: Efficient Wavelet-Transformer for single image denoising.

Transformer-based image denoising methods have shown remarkable potential but suffer from high computational cost and large memory footprint due to their linear operations for capturing long-range dependencies. In this work, we aim to develop a more resource-efficient Transformer-based image denoising method that maintains high performance. To this end, we propose an Efficient Wavelet Transformer (EWT), which incorporates a Frequency-domain Conversion Pipeline (FCP) to reduce image resolution without losing critical features, and a Multi-level Feature Aggregation Module (MFAM) with a Dual-stream Feature Extraction Block (DFEB) to harness hierarchical features effectively. EWT achieves a faster processing speed by over 80% and reduces GPU memory usage by more than 60% compared to the original Transformer, while still delivering denoising performance on par with state-of-the-art methods. Extensive experiments show that EWT significantly improves the efficiency of Transformer-based image denoising, providing a more balanced approach between performance and resource consumption.

Expression, purification, and crystal structure of mpox virus A41 protein.

Mpox is a zoonotic disease that was once endemic in Africa countries caused by mpox virus. However, cases recently have been confirmed in many non-endemic countries outside of Africa. The rapidly increasing number of confirmed mpox cases poses a threat to the international community. In-depth studies of key viral factors are urgently needed, which will inform the design of multiple antiviral agents. Mpox virus A41L gene encodes a secreted protein, A41, that is nonessential for viral replication, but could affect the host response to infection via interacting with chemokines. Here, mpox virus A41 protein was expressed in Sf9 cells, and purified by affinity chromatography followed by gel filtration. Surface plasmon resonance spectroscopy showed that purified A41 binds a certain human chemokine CXCL8 with the equilibrium dissociation constant (KD) being 1.22 × 10-6 M. The crystal structure of mpox virus A41 protein was solved at 1.92 Å. Structural analysis and comparison revealed that mpox virus A41 protein adopts a characteristic ß-sheet topology, showing minor differences with that of vaccinia virus. These preliminary structural and functional studies of A41 protein from mpox virus will help us better understand its role in chemokine subversion, and contributing to the knowledge to viral chemokine binding proteins.

Involution Transformer based U-Net for Landmark Detection in Ultrasound Images for Diagnosis of Infantile DDH.

The B-mode ultrasound based computer-aided diagnosis (CAD) has demonstrated its effectiveness for diagnosis of Developmental Dysplasia of the Hip (DDH) in infants, which can conduct the Graf's method by detecting landmarks in hip ultrasound images. However, it is still necessary to explore more valuable information around these landmarks to enhance feature representation for improving detection performance in the detection model. To this end, a novel Involution Transformer based U-Net (IT-UNet) network is proposed for hip landmark detection. The IT-UNet integrates the efficient involution operation into Transformer to develop an Involution Transformer module (ITM), which consists of an involution attention block and a squeeze-and-excitation involution block. The ITM can capture both the spatial-related information and long-range dependencies from hip ultrasound images to effectively improve feature representation. Moreover, an Involution Downsampling block (IDB) is developed to alleviate the issue of feature loss in the encoder modules, which combines involution and convolution for the purpose of downsampling. The experimental results on two DDH ultrasound datasets indicate that the proposed IT-UNet achieves the best landmark detection performance, indicating its potential applications.

Cost-Sensitive Weighted Contrastive Learning Based on Graph Convolutional Networks for Imbalanced Alzheimer's Disease Staging.

Identifying the progression stages of Alzheimer's disease (AD) can be considered as an imbalanced multi-class classification problem in machine learning. It is challenging due to the class imbalance issue and the heterogeneity of the disease. Recently, graph convolutional networks (GCNs) have been successfully applied in AD classification. However, these works did not handle the class imbalance issue in classification. Besides, they ignore the heterogeneity of the disease. To this end, we propose a novel cost-sensitive weighted contrastive learning method based on graph convolutional networks (CSWCL-GCNs) for imbalanced AD staging using resting-state functional magnetic resonance imaging (rs-fMRI). The proposed method is developed on a multi-view graph constructed using the functional connectivity (FC) and high-order functional connectivity (HOFC) features of the subjects. A novel cost-sensitive weighted contrastive learning procedure is proposed to capture discriminative information from the minority classes, encouraging the samples in the minority class to provide adequate supervision. Considering the heterogeneity of the disease, the weights of the negative pairs are introduced into contrastive learning and they are computed based on the distance to class prototypes, which are automatically learned from the training data. Meanwhile, the cost-sensitive mechanism is further introduced into contrastive learning to handle the class imbalance issue. The proposed CSWCL-GCN is evaluated on 720 subjects (including 184 NCs, 40 SMC patients, 208 EMCI patients, 172 LMCI patients and 116 AD patients) from the ADNI (Alzheimer's Disease Neuroimaging Initiative). Experimental results show that the proposed CSWCL-GCN outperforms state-of-the-art methods on the ADNI database.

Weakly Supervised Lesion Detection and Diagnosis for Breast Cancers with Partially Annotated Ultrasound Images.

Deep learning (DL) has proven highly effective for ultrasound-based computer-aided diagnosis (CAD) of breast cancers. In an automatic CAD system, lesion detection is critical for the following diagnosis. However, existing DL-based methods generally require voluminous manually-annotated region of interest (ROI) labels and class labels to train both the lesion detection and diagnosis models. In clinical practice, the ROI labels, i.e. ground truths, may not always be optimal for the classification task due to individual experience of sonologists, resulting in the issue of coarse annotation to limit the diagnosis performance of a CAD model. To address this issue, a novel Two-Stage Detection and Diagnosis Network (TSDDNet) is proposed based on weakly supervised learning to improve diagnostic accuracy of the ultrasound-based CAD for breast cancers. In particular, all the initial ROI-level labels are considered as coarse annotations before model training. In the first training stage, a candidate selection mechanism is then designed to refine manual ROIs in the fully annotated images and generate accurate pseudo-ROIs for the partially annotated images under the guidance of class labels. The training set is updated with more accurate ROI labels for the second training stage. A fusion network is developed to integrate detection network and classification network into a unified end-to-end framework as the final CAD model in the second training stage. A self-distillation strategy is designed on this model for joint optimization to further improves its diagnosis performance. The proposed TSDDNet is evaluated on three B-mode ultrasound datasets, and the experimental results indicate that it achieves the best performance on both lesion detection and diagnosis tasks, suggesting promising application potential.

A fine-grained orthodontics segmentation model for 3D intraoral scan data.

With the widespread application of digital orthodontics in the diagnosis and treatment of oral diseases, more and more researchers focus on the accurate segmentation of teeth from intraoral scan data. The accuracy of the segmentation results will directly affect the follow-up diagnosis of dentists. Although the current research on tooth segmentation has achieved promising results, the 3D intraoral scan datasets they use are almost all indirect scans of plaster models, and only contain limited samples of abnormal teeth, so it is difficult to apply them to clinical scenarios under orthodontic treatment. The current issue is the lack of a unified and standardized dataset for analyzing and validating the effectiveness of tooth segmentation. In this work, we focus on deformed teeth segmentation and provide a fine-grained tooth segmentation dataset (3D-IOSSeg). The dataset consists of 3D intraoral scan data from more than 200 patients, with each sample labeled with a fine-grained mesh unit. Meanwhile, 3D-IOSSeg meticulously classified every tooth in the upper and lower jaws. In addition, we propose a fast graph convolutional network for 3D tooth segmentation named Fast-TGCN. In the model, the relationship between adjacent mesh cells is directly established by the naive adjacency matrix to better extract the local geometric features of the tooth. Extensive experiments show that Fast-TGCN can quickly and accurately segment teeth from the mouth with complex structures and outperforms other methods in various evaluation metrics. Moreover, we present the results of multiple classical tooth segmentation methods on this dataset, providing a comprehensive analysis of the field. All code and data will be available at https://github.com/MIVRC/Fast-TGCN.

Pseudo-Data Based Self-Supervised Federated Learning for Classification of Histopathological Images.

Computer-aided diagnosis (CAD) can help pathologists improve diagnostic accuracy together with consistency and repeatability for cancers. However, the CAD models trained with the histopathological images only from a single center (hospital) generally suffer from the generalization problem due to the straining inconsistencies among different centers. In this work, we propose a pseudo-data based self-supervised federated learning (FL) framework, named SSL-FT-BT, to improve both the diagnostic accuracy and generalization of CAD models. Specifically, the pseudo histopathological images are generated from each center, which contain both inherent and specific properties corresponding to the real images in this center, but do not include the privacy information. These pseudo images are then shared in the central server for self-supervised learning (SSL) to pre-train the backbone of global mode. A multi-task SSL is then designed to effectively learn both the center-specific information and common inherent representation according to the data characteristics. Moreover, a novel Barlow Twins based FL (FL-BT) algorithm is proposed to improve the local training for the CAD models in each center by conducting model contrastive learning, which benefits the optimization of the global model in the FL procedure. The experimental results on four public histopathological image datasets indicate the effectiveness of the proposed SSL-FL-BT on both diagnostic accuracy and generalization.

Heterobifunctional Cross-Linker with Dinitroimidazole and N-Hydroxysuccinimide Ester Motifs for Protein Functionalization and Cysteine-Lysine Peptide Stapling.

A heterobifunctional cross-linker with one sulfhydryl-reactive dinitroimidazole end and another amine-reactive N-hydroxysuccinimide (NHS) ester end was designed and synthesized. The two motifs of this cross-linker, dinitroimidazole and NHS ester, proved to react with thiol and amine, respectively, in an orthogonal way. The cross-linker was further applied to construct stapled peptides of different sizes and mono- and dual functionalization (including biotinylation, PEGylation, and fluorescence labeling) of protein.

Recent Advances in Application of Alkoxy Radical in Organic Synthesis.

Alkoxy radicals have been identified as versatile intermediates in synthetic chemistry in the last few decades. Over the last decade, various catalytic processes for the in situ generation of alkoxy radicals have been explored, leading to the development of new synthetic methodologies based on alkoxy radicals. In this review, we provided a comprehensive review of recent developments in the utilization of alkoxy radicals in diverse organic transformations, natural product synthesis, and the late-stage modification of bioactive molecules through the implementation of the photoredox methodology.

How does blockchain technology affect the development of green finance? Theoretical analysis and empirical verification.

Facing the problem of a lack of endogenous incentive mechanisms for the development of green finance, we regard blockchain technology as an institutional technology and elevate it to the height of governance mechanisms. Using a dynamic stochastic general equilibrium (DSGE) model framework, we compare and analyze its effects with traditional supportive policies such as fiscal subsidies. The modeling simulation results show that the blockchain green finance platform model is conducive to better promoting the development of green finance. Subsequently, we construct a financial technology development index centered on blockchain technology and empirically test the impact of blockchain financial technology on the level of green finance development from both the supply and demand sides. The results show that the development of blockchain financial technology has significantly increased the scale of green credit issuance and effectively eased the financing constraints of green enterprises, reducing financing costs. We conduct an economic analysis of the impact of blockchain financial technology on the development of green finance, providing a feasible path for the integration and development of green finance and financial technology.

Coexistence of primary thyroid diffuse large B cell lymphoma and papillary thyroid carcinoma in a patient with Hashimoto's thyroiditis: a case report and literature review.

Papillary thyroid carcinoma (PTC) is the most common pathological type of thyroid malignancy and also has an excellent prognosis. Primary thyroid lymphoma (PTL) is rare and has a poor prognosis. The co-occurrence of both malignancies is extremely rare, and the preoperative diagnosis is rather difficult. We report the case of a patient with both PTC and PTL in the setting of Hashimoto's thyroiditis (HT). A 59-year-old female patient was referred to our department for progressive enlargement of the thyroid gland over a few months. The imaging results demonstrated an enlarged thyroid and a mass in the thyroid. Total thyroidectomy and bilateral central neck node dissection were conducted. The final diagnosis of the coexistence of thyroid diffuse large B cell lymphoma and PTC was confirmed by histopathology and immunohistochemistry. The patient received radiation therapy and six cycles of chemotherapy combined with targeted therapy, including rituximab, cyclophosphamide, doxorubicin, vindesine, and prednisone (R-CHOP). After 6 months of follow-up, neither tumor has recurred. It is important for physicians to keep PTL in mind for differential diagnosis in HT patients with sudden thyroid enlargement.

Multi-OMIC analysis of Huntington disease reveals a neuroprotective astrocyte state.

Huntington disease (HD) is an incurable neurodegenerative disease characterized by neuronal loss and astrogliosis. One hallmark of HD is the selective neuronal vulnerability of striatal medium spiny neurons. To date, the underlying mechanisms of this selective vulnerability have not been fully defined. Here, we employed a multi-omic approach including single nucleus RNAseq (snRNAseq), bulk RNAseq, lipidomics, HTT gene CAG repeat length measurements, and multiplexed immunofluorescence on post-mortem brain tissue from multiple brain regions of HD and control donors. We defined a signature of genes that is driven by CAG repeat length and found it enriched in astrocytic and microglial genes. Moreover, weighted gene correlation network analysis showed loss of connectivity of astrocytic and microglial modules in HD and identified modules that correlated with CAG-repeat length which further implicated inflammatory pathways and metabolism. We performed lipidomic analysis of HD and control brains and identified several lipid species that correlate with HD grade, including ceramides and very long chain fatty acids. Integration of lipidomics and bulk transcriptomics identified a consensus gene signature that correlates with HD grade and HD lipidomic abnormalities and implicated the unfolded protein response pathway. Because astrocytes are critical for brain lipid metabolism and play important roles in regulating inflammation, we analyzed our snRNAseq dataset with an emphasis on astrocyte pathology. We found two main astrocyte types that spanned multiple brain regions; these types correspond to protoplasmic astrocytes, and fibrous-like - CD44-positive, astrocytes. HD pathology was differentially associated with these cell types in a region-specific manner. One protoplasmic astrocyte cluster showed high expression of metallothionein genes, the depletion of this cluster positively correlated with the depletion of vulnerable medium spiny neurons in the caudate nucleus. We confirmed that metallothioneins were increased in cingulate HD astrocytes but were unchanged or even decreased in caudate astrocytes. We combined existing genome-wide association studies (GWAS) with a GWA study conducted on HD patients from the original Venezuelan cohort and identified a single-nucleotide polymorphism in the metallothionein gene locus associated with delayed age of onset. Functional studies found that metallothionein overexpressing astrocytes are better able to buffer glutamate and were neuroprotective of patient-derived directly reprogrammed HD MSNs as well as against rotenone-induced neuronal death in vitro. Finally, we found that metallothionein-overexpressing astrocytes increased the phagocytic activity of microglia in vitro and increased the expression of genes involved in fatty acid binding. Together, we identified an astrocytic phenotype that is regionally-enriched in less vulnerable brain regions that can be leveraged to protect neurons in HD.

The role of programmed cell death in diabetic foot ulcers.

Diabetic foot ulcer, is a chronic complication afflicting individuals with diabetes, continue to increase worldwide, immensely burdening society. Programmed cell death, which includes apoptosis, autophagy, ferroptosis, necroptosis and pyroptosis, has been increasingly implicated in the pathogenesis of diabetic foot ulcer. This review is based on an exhaustive examination of the literature on 'programmed cell death' and 'diabetic foot ulcers' via PubMed. The findings revealed that natural bioactive compounds, noncoding RNAs and certain proteins play crucial roles in the healing of diabetic foot ulcers through various forms of programmed cell death, including apoptosis, autophagy, ferroptosis and pyroptosis.

Multi-Scale Efficient Graph-Transformer for Whole Slide Image Classification.

The multi-scale information among the whole slide images (WSIs) is essential for cancer diagnosis. Although the existing multi-scale vision Transformer has shown its effectiveness for learning multi-scale image representation, it still cannot work well on the gigapixel WSIs due to their extremely large image sizes. To this end, we propose a novel Multi-scale Efficient Graph-Transformer (MEGT) framework for WSI classification. The key idea of MEGT is to adopt two independent efficient Graph-based Transformer (EGT) branches to process the low-resolution and high-resolution patch embeddings (i.e., tokens in a Transformer) of WSIs, respectively, and then fuse these tokens via a multi-scale feature fusion module (MFFM). Specifically, we design an EGT to efficiently learn the local-global information of patch tokens, which integrates the graph representation into Transformer to capture spatial-related information of WSIs. Meanwhile, we propose a novel MFFM to alleviate the semantic gap among different resolution patches during feature fusion, which creates a non-patch token for each branch as an agent to exchange information with another branch by cross-attention mechanism. In addition, to expedite network training, a new token pruning module is developed in EGT to reduce the redundant tokens. Extensive experiments on both TCGA-RCC and CAMELYON16 datasets demonstrate the effectiveness of the proposed MEGT.

Comparative Efficacy and Acceptability of 12 Phosphorus-Lowering Drugs in Adults with Hyperphosphatemia and Chronic Kidney Disease: A Systematic Review and Network Meta-Analysis.

BACKGROUND: Hyperphosphatemia is associated with cardiovascular morbidity and mortality in adults with chronic kidney disease (CKD). Drug therapy has an irreplaceable role in the management of hyperphosphatemia. OBJECTIVES: We aimed to compare and rank phosphorus-lowering drugs, including phosphate binder and nonphosphate binder, in hyperphosphatemia adults with CKD. METHODS: We did a systematic review and frequentist random-effect network meta-analysis. We searched in PubMed, Cochrane Library, Web of Science, and Embase from inception to February 1, 2023, for randomized controlled trials of 12 phosphorus-lowering drugs in adults with hyperphosphatemia and CKD. Primary outcomes were efficacy (changes in serum phosphorus) and acceptability (treatment withdrawals due to any cause). We ranked each drug according to the value of surface under the cumulative ranking curve. We applied the Confidence in Network Meta-Analysis frameworks to rate the certainty of evidence. This study was registered with PROSPERO, number CRD42022322270. RESULTS: We identified 2,174 citations, and of these, we included 94 trials comprising 14,459 participants and comparing 13 drugs or placebo. In terms of efficacy, except for niacinamide, all drugs lowered the level of serum phosphorus compared with placebo, with mean difference ranging between -1.61 (95% credible interval [CrI], -2.60 to -0.62) mg/dL for magnesium carbonate and -0.85 (-1.66 to -0.05) mg/dL for bixalomer. Only ferric citrate with odds ratios 0.56 (95% CrI: 0.36-0.89) was significantly associated with fewer dropouts for acceptability. Of the 94 trials, 43 (46%), 7 (7%), and 44 (47%) trials were rated as high, moderate, and low risk of bias, respectively, the certainty of the evidence was moderate to very low. CONCLUSIONS: Magnesium carbonate has the best phosphorus-lowering effect in hyperphosphatemia adults with CKD; considering efficacy and acceptability, ferric citrate shows evidence to be the most appropriate drug with or without dialysis.

Hounsfield units on abdominal computed tomography: a new tool for predicting osteoporosis.

BACKGROUND: Osteoporosis can cause bone fractures and disability, but early diagnosis faces challenges. Our proposed diagnostic indicators offer a new approach for early detection, which benefits early identification. PURPOSE: To determine the most appropriate threshold for predicting osteoporosis in patients with each section of vertebral body. MATERIAL AND METHODS: A retrospective analysis of 210 patients, including 646 vertebrae, who had both abdominal computed tomography (CT) and dual-energy X-ray absorptiometry (DXA) within six months. The correlation between DXA T-score and CT Hounsfield units (HU) values was tested by Pearson. The area under the curve (AUC) was calculated using the threshold obtained from the regression equation. RESULTS: The thresholds matching the T-score of -2.5 were 85, 95, 85, and 90 HU for the upper axial plane of the vertebral body (Lau), the middle axial plane of the vertebral body (Lam), the lower axial plane of the vertebral body (Lad), and the mid-sagittal plane of the vertebral body (Lsm), respectively. Defining osteoporosis using CT as Lau ≤ 85, Lam ≤ 95, Lad ≤ 85, or Lsm ≤ 90 HU had a specificity of 88.1% (116/134) and sensitivity of 90.8% (69/76) for distinguishing DXA osteoporosis of the lumbar spine in 210 patients. T-score ≤-2.5 defined as Lau ≤85 or Lam ≤95 or Lad ≤85 or Lsm ≤90 HU had a specificity of 85.9% (275/320) and sensitivity of 82.8% (270/326) for DXA T-score ≤-2.5 in 646 lumbar vertebrae. CONCLUSION: CT HU values obtained based on different sections of the vertebral body in abdominal CT can be used as a supplementary measure to assess osteoporosis.

The link between SARS-CoV-2 related microglial reactivity and astrocyte pathology in the inferior olivary nucleus.

The pathological involvement of the central nervous system in SARS-CoV2 (COVID-19) patients is established. The burden of pathology is most pronounced in the brain stem including the medulla oblongata. Hypoxic/ischemic damage is the most frequent neuropathologic abnormality. Other neuropathologic features include neuronophagia, microglial nodules, and hallmarks of neurodegenerative diseases: astrogliosis and microglial reactivity. It is still unknown if these pathologies are secondary to hypoxia versus a combination of inflammatory response combined with hypoxia. It is also unknown how astrocytes react to neuroinflammation in COVID-19, especially considering evidence supporting the neurotoxicity of certain astrocytic phenotypes. This study aims to define the link between astrocytic and microglial pathology in COVID-19 victims in the inferior olivary nucleus, which is one of the most severely affected brain regions in COVID-19, and establish whether COVID-19 pathology is driven by hypoxic damage. Here, we conducted neuropathologic assessments and multiplex-immunofluorescence studies on the medulla oblongata of 18 COVID-19, 10 pre-pandemic patients who died of acute respiratory distress syndrome (ARDS), and 7-8 control patients with no ARDS or COVID-19. The comparison of ARDS and COVID-19 allows us to identify whether the pathology in COVID-19 can be explained by hypoxia alone, which is common to both conditions. Our results showed increased olivary astrogliosis in ARDS and COVID-19. However, microglial density and microglial reactivity were increased only in COVID-19, in a region-specific manner. Also, olivary hilar astrocytes increased YKL-40 (CHI3L1) in COVID-19, but to a lesser extent than ARDS astrocytes. COVID-19 astrocytes also showed lower levels of Aquaporin-4 (AQP4), and Metallothionein-3 in subsets of COVID-19 brain regions. Cluster analysis on immunohistochemical attributes of astrocytes and microglia identified ARDS and COVID-19 clusters with correlations to clinical history and disease course. Our results indicate that olivary glial pathology and neuroinflammation in the COVID-19 cannot be explained solely by hypoxia and suggest that failure of astrocytes to upregulate the anti-inflammatory YKL-40 may contribute to the neuroinflammation. Notwithstanding the limitations of retrospective studies in establishing causality, our experimental design cannot adequately control for factors external to our design. Perturbative studies are needed to confirm the role of the above-described astrocytic phenotypes in neuroinflammation.

Efficacy and safety of tenapanor in hemodialysis patients with hyperphosphatemia: A systematic review and meta-analysis of randomized placebo-controlled trials.

BACKGROUND: The effects of tenapanor in reducing serum phosphorus in hemodialysis patients with hyperphosphatemia are uncertain and no relevant meta-analysis has been conducted. We performed a meta-analysis of randomized placebo-controlled trials to evaluate the efficacy and safety of tenapanor. METHODS: All randomized controlled trials of tenapanor were searched up to 1 August 2022. The primary endpoint was the change in serum phosphorus level from baseline with tenapanor and placebo. Data on drug-related adverse events (AEs), gastrointestinal AEs and diarrhea were collected to determine the safety of tenapanor. RESULTS: There were 533 patients throughout five trials that were eligible. Tenapanor significantly lowered blood phosphorus level by 1.79 mg/dl in the mean difference than the placebo. Diarrhea, gastrointestinal AEs, and drug-related AEs were more severe than placebo. CONCLUSIONS: This meta-analysis showed that although drug side effects were common, tenapanor significantly reduced serum phosphorus level in hemodialysis patients.

Variational Cross-Graph Reasoning and Adaptive Structured Semantics Learning for Compositional Temporal Grounding.

Temporal grounding is the task of locating a specific segment from an untrimmed video according to a query sentence. This task has achieved significant momentum in the computer vision community as it enables activity grounding beyond pre-defined activity classes by utilizing the semantic diversity of natural language descriptions. The semantic diversity is rooted in the principle of compositionality in linguistics, where novel semantics can be systematically described by combining known words in novel ways (compositional generalization). However, existing temporal grounding datasets are not carefully designed to evaluate the compositional generalizability. To systematically benchmark the compositional generalizability of temporal grounding models, we introduce a new Compositional Temporal Grounding task and construct two new dataset splits, i.e., Charades-CG and ActivityNet-CG. We empirically find that they fail to generalize to queries with novel combinations of seen words. We argue that the inherent compositional structure (i.e., composition constituents and their relationships) inside the videos and language is the crucial factor to achieve compositional generalization. Based on this insight, we propose a variational cross-graph reasoning framework that explicitly decomposes video and language into hierarchical semantic graphs, respectively, and learns fine-grained semantic correspondence between the two graphs. Meanwhile, we introduce a novel adaptive structured semantics learning approach to derive the structure-informed and domain-generalizable graph representations, which facilitate the fine-grained semantic correspondence reasoning between the two graphs. To further evaluate the understanding of the compositional structure, we also introduce a more challenging setting, where one of the components in the novel composition is unseen. This requires more sophisticated understanding of the compositional structure to infer the potential semantics of the unseen word based on the other learned composition constituents appearing in both the video and language context, and their relationships. Extensive experiments validate the superior compositional generalizability of our approach, demonstrating its ability to handle queries with novel combinations of seen words as well as novel words in the testing composition.

Re-convolving the compositional landscape of primary and recurrent glioblastoma reveals prognostic and targetable tissue states.

Glioblastoma (GBM) diffusely infiltrates the brain and intermingles with non-neoplastic brain cells, including astrocytes, neurons and microglia/myeloid cells. This complex mixture of cell types forms the biological context for therapeutic response and tumor recurrence. We used single-nucleus RNA sequencing and spatial transcriptomics to determine the cellular composition and transcriptional states in primary and recurrent glioma and identified three compositional 'tissue-states' defined by cohabitation patterns between specific subpopulations of neoplastic and non-neoplastic brain cells. These tissue-states correlated with radiographic, histopathologic, and prognostic features and were enriched in distinct metabolic pathways. Fatty acid biosynthesis was enriched in the tissue-state defined by the cohabitation of astrocyte-like/mesenchymal glioma cells, reactive astrocytes, and macrophages, and was associated with recurrent GBM and shorter survival. Treating acute slices of GBM with a fatty acid synthesis inhibitor depleted the transcriptional signature of this pernicious tissue-state. These findings point to therapies that target interdependencies in the GBM microenvironment.