M2ixKG: Mixing for harder negative samples in knowledge graph.

Knowledge graph embedding (KGE) involves mapping entities and relations to low-dimensional dense embeddings, enabling a wide range of real-world applications. The mapping is achieved via distinguishing the positive and negative triplets in knowledge graphs. Therefore, how to design high-quality negative triplets is critical in the effectiveness of KEG models. Existing KGE models face challenges in generating high-quality negative triplets. Some models employ simple static distributions, i.e. uniform or Bernoulli distribution, and it is difficult for these methods to be trained distinguishably because of the sampled uninformative negative triplets. Furthermore, current methods are confined to constructing negative triplets from existing entities within the knowledge graph, limiting their ability to explore harder negatives. We introduce a novel mixing strategy in knowledge graphs called M2ixKG. M2ixKG adopts mixing operation in generating harder negative samples from two aspects: one is mixing among the heads and tails in triplets with the same relation to strengthen the robustness and generalization of the entity embeddings; the other is mixing the negatives with high scores to generate harder negatives. Our experiments, utilizing three datasets and four classical score functions, highlight the exceptional performance of M2ixKG in comparison to previous negative sampling algorithms.

Bayesian hypernetwork collaborates with time-difference evolutional network for temporal knowledge prediction.

A Temporal Knowledge Graph (TKG) is a sequence of Knowledge Graphs (KGs) attached with time information, in which each KG contains the facts that co-occur at the same timestamp. Temporal knowledge prediction (TKP) aims to predict future events given observed historical KGs in TKGs, which is essential for many applications to provide intelligent analysis services. However, most existing TKP methods focus on entity and relation prediction tasks but ignore the importance of time prediction tasks. Furthermore, there is uncertainty in time prediction, and it is difficult for prediction models to model it completely. In this work, we propose a collaboration framework with Bayesian Hypernetwork and Time-Difference Evolutional Network (BH-TDEN) to address these problems. First, we begin with the time prediction task, and we present a Bayesian hypernetwork to model the uncertainty of events time. For the input of Bayesian hypernetwork, we design a novel time-difference evolutional network to obtain the entities and relations embedding. Specifically, we propose an auto-regressive time gate parameterized by the time difference of adjacent KGs in entity and relation encoder to learn the time-sensitive TKG embedding, which not only learns the relationship between the given time information and TKG embedding but also provides more expressive TKG embedding for Bayesian hypernetwork to accurately predict the time of future events. Furthermore, we also present a novel relation updating mechanism that employs the neighbor relations of the subject corresponding to the current relation to learn more adaptive relation embedding. Extensive experiments demonstrate that the proposed method obtains considerable time prediction and link prediction performance on four TKG benchmark datasets.

Spatial reconstructed local attention Res2Net with F0 subband for fake speech detection.

The rhythm of bonafide speech is often difficult to replicate, which causes that the fundamental frequency (F0) of synthetic speech is significantly different from that of real speech. It is expected that the F0 feature contains the discriminative information for the fake speech detection (FSD) task. In this paper, we propose a novel F0 subband for FSD. In addition, to effectively model the F0 subband so as to improve the performance of FSD, the spatial reconstructed local attention Res2Net (SR-LA Res2Net) is proposed. Specifically, Res2Net is used as a backbone network to obtain multiscale information, and enhanced with a spatial reconstruction mechanism to avoid losing important information when the channel group is constantly superimposed. In addition, local attention is designed to make the model focus on the local information of the F0 subband. Experimental results on the ASVspoof 2019 LA dataset show that our proposed method obtains an equal error rate (EER) of 0.47% and a minimum tandem detection cost function (min t-DCF) of 0.0159, achieving the state-of-the-art performance among all of the single systems.

A systems biology approach uncovers novel disease mechanisms in age-related macular degeneration.

Age-related macular degeneration (AMD) is a leading cause of blindness, affecting 200 million people worldwide. To identify genes that could be targeted for treatment, we created a molecular atlas at different stages of AMD. Our resource is comprised of RNA sequencing (RNA-seq) and DNA methylation microarrays from bulk macular retinal pigment epithelium (RPE)/choroid of clinically phenotyped normal and AMD donor eyes (n = 85), single-nucleus RNA-seq (164,399 cells), and single-nucleus assay for transposase-accessible chromatin (ATAC)-seq (125,822 cells) from the retina, RPE, and choroid of 6 AMD and 7 control donors. We identified 23 genome-wide significant loci differentially methylated in AMD, over 1,000 differentially expressed genes across different disease stages, and an AMD Müller state distinct from normal or gliosis. Chromatin accessibility peaks in genome-wide association study (GWAS) loci revealed putative causal genes for AMD, including HTRA1 and C6orf223. Our systems biology approach uncovered molecular mechanisms underlying AMD, including regulators of WNT signaling, FRZB and TLE2, as mechanistic players in disease.

GCNet: Graph Completion Network for Incomplete Multimodal Learning in Conversation.

Conversations have become a critical data format on social media platforms. Understanding conversation from emotion, content and other aspects also attracts increasing attention from researchers due to its widespread application in human-computer interaction. In real-world environments, we often encounter the problem of incomplete modalities, which has become a core issue of conversation understanding. To address this problem, researchers propose various methods. However, existing approaches are mainly designed for individual utterances rather than conversational data, which cannot fully exploit temporal and speaker information in conversations. To this end, we propose a novel framework for incomplete multimodal learning in conversations, called "Graph Complete Network (GCNet)," filling the gap of existing works. Our GCNet contains two well-designed graph neural network-based modules, "Speaker GNN" and "Temporal GNN," to capture temporal and speaker dependencies. To make full use of complete and incomplete data, we jointly optimize classification and reconstruction tasks in an end-to-end manner. To verify the effectiveness of our method, we conduct experiments on three benchmark conversational datasets. Experimental results demonstrate that our GCNet is superior to existing state-of-the-art approaches in incomplete multimodal learning.

Adaptive pseudo-Siamese policy network for temporal knowledge prediction.

Temporal knowledge prediction is a crucial task for early event warning, which has gained increasing attention recently. It aims to predict future facts based on relevant historical facts using temporal knowledge graphs. There are two main difficulties associated with the prediction task: from the perspective of historical facts, modeling the evolutionary patterns of facts to accurately predict the query and from the query perspective, handling the two cases where the query contains seen and unseen entities in a unified framework. Driven by these two problems, we propose a novel adaptive pseudo-Siamese policy network for temporal knowledge prediction based on reinforcement learning. Specifically, we design the policy network in our model as a pseudo-Siamese network consisting of two sub-policy networks. In the sub-policy network I, the agent searches for the answer to the query along the entity-relation paths to capture static evolutionary patterns. In sub-policy network II, the agent searches for the answer to the query along relation-time paths to deal with unseen entities. Moreover, we develop a temporal relation encoder to capture the temporal evolutionary patterns. Finally, we design a gating mechanism to adaptively integrate the results of the two sub-policy networks to help the agent focus on the destination answer. To assess the performance of our model, we conduct link prediction on four benchmark datasets, and extensive experimental results demonstrate that our method achieves considerable performance compared with existing methods.

PIRNet: Personality-Enhanced Iterative Refinement Network for Emotion Recognition in Conversation.

Emotion recognition in conversation (ERC) is important for enhancing user experience in human-computer interaction. Unlike vanilla emotion recognition in individual utterances, ERC aims to classify constituent utterances in a dialog into corresponding emotion labels, which makes contextual information crucial. In addition to contextual information, personality traits also affect emotional perception based on psychological findings. Although researchers have proposed several approaches and achieved promising results on ERC, current works in this domain rarely incorporate contextual information and personality influence. To this end, we propose a novel framework to integrate these factors seamlessly, called "Personality-enhanced Iterative Refinement Network (PIRNet)." Specifically, PIRNet is a multistage iterative method. To capture personality influence, PIRNet leverages personality traits to mimic emotional transitions and generates personality-enhanced results. Then we exploit sequence models to capture contextual information in conversations. To verify the effectiveness of our proposed method, we conduct experiments on three benchmark datasets for ERC, that is, IEMOCAP, CMU-MOSI, and CMU-MOSEI. Experimental results demonstrate that our PIRNet succeeds over currently advanced approaches to emotion recognition.

Preoperative Prediction of Malignant Transformation of Sinonasal Inverted Papilloma Using MR Radiomics.

Purpose: Accurate preoperative prediction of the malignant transformation of sinonasal inverted papilloma (IP) is essential for guiding biopsy, planning appropriate surgery and prognosis of patients. We aimed to investigate the value of MRI-based radiomics in discriminating IP from IP-transformed squamous cell carcinomas (IP-SCC). Methods: A total of 236 patients with IP-SCC (n=92) or IP (n=144) were enrolled and divided into a training cohort and a testing cohort. Preoperative MR images including T1-weighted, T2-weighted, and contrast enhanced T1-weighted images were collected. Radiomic features were extracted from MR images and key features were merged into a radiomic model. A morphological features model was developed based on MR morphological features assessed by radiologists. A combined model combining radiomic features and morphological features was generated using multivariable logistic regression. For comparison, two head and neck radiologists were independently invited to distinguish IP-SCC from IP. The area under the receiver operating characteristics curve (AUC) was used to assess the performance of all models. Results: A total of 3948 radiomic features were extracted from three MR sequences. After feature selection, we saved 15 key features for modeling. The AUC, sensitivity, specificity, and accuracy on the testing cohort of the combined model based on radiomic and morphological features were respectively 0.962, 0.828, 0.94, and 0.899. The diagnostic ability of the combined model outperformed the morphological features model and also outperformed the two head and neck radiologists. Conclusions: A combined model based on MR radiomic and morphological features could serve as a potential tool to accurately predict IP-SCC, which might improve patient counseling and make more precise treatment planning.

Combining a parallel 2D CNN with a self-attention Dilated Residual Network for CTC-based discrete speech emotion recognition.

A challenging issue in the field of the automatic recognition of emotion from speech is the efficient modelling of long temporal contexts. Moreover, when incorporating long-term temporal dependencies between features, recurrent neural network (RNN) architectures are typically employed by default. In this work, we aim to present an efficient deep neural network architecture incorporating Connectionist Temporal Classification (CTC) loss for discrete speech emotion recognition (SER). Moreover, we also demonstrate the existence of further opportunities to improve SER performance by exploiting the properties of convolutional neural networks (CNNs) when modelling contextual information. Our proposed model uses parallel convolutional layers (PCN) integrated with Squeeze-and-Excitation Network (SEnet), a system herein denoted as PCNSE, to extract relationships from 3D spectrograms across timesteps and frequencies; here, we use the log-Mel spectrogram with deltas and delta-deltas as input. In addition, a self-attention Residual Dilated Network (SADRN) with CTC is employed as a classification block for SER. To the best of the authors' knowledge, this is the first time that such a hybrid architecture has been employed for discrete SER. We further demonstrate the effectiveness of our proposed approach on the Interactive Emotional Dyadic Motion Capture (IEMOCAP) and FAU-Aibo Emotion corpus (FAU-AEC). Our experimental results reveal that the proposed method is well-suited to the task of discrete SER, achieving a weighted accuracy (WA) of 73.1% and an unweighted accuracy (UA) of 66.3% on IEMOCAP, as well as a UA of 41.1% on the FAU-AEC dataset.

Apolipoprotein L1-Specific Antibodies Detect Endogenous APOL1 inside the Endoplasmic Reticulum and on the Plasma Membrane of Podocytes.

BACKGROUND: APOL1 is found in human kidney podocytes and endothelia. Variants G1 and G2 of the APOL1 gene account for the high frequency of nondiabetic CKD among African Americans. Proposed mechanisms of kidney podocyte cytotoxicity resulting from APOL1 variant overexpression implicate different subcellular compartments. It is unclear where endogenous podocyte APOL1 resides, because previous immunolocalization studies utilized overexpressed protein or commercially available antibodies that crossreact with APOL2. This study describes and distinguishes the locations of both APOLs. METHODS: Immunohistochemistry, confocal and immunoelectron microscopy, and podocyte fractionation localized endogenous and transfected APOL1 using a large panel of novel APOL1-specific mouse and rabbit monoclonal antibodies. RESULTS: Both endogenous podocyte and transfected APOL1 isoforms vA and vB1 (and a little of isoform vC) localize to the luminal face of the endoplasmic reticulum (ER) and to the cell surface, but not to mitochondria, endosomes, or lipid droplets. In contrast, APOL2, isoform vB3, and most vC of APOL1 localize to the cytoplasmic face of the ER and are consequently absent from the cell surface. APOL1 knockout podocytes do not stain for APOL1, attesting to the APOL1-specificity of the antibodies. Stable re-transfection of knockout podocytes with inducible APOL1-G0, -G1, and -G2 showed no differences in localization among variants. CONCLUSIONS: APOL1 is found in the ER and plasma membrane, consistent with either the ER stress or surface cation channel models of APOL1-mediated cytotoxicity. The surface localization of APOL1 variants potentially opens new therapeutic targeting avenues.

Publisher Correction: Classical and alternative complement activation on photoreceptor outer segments drives monocyte-dependent retinal atrophy.

A correction to this article has been published and is linked from the HTML and PDF versions of this paper. The error has been fixed in the paper.

Classical and alternative complement activation on photoreceptor outer segments drives monocyte-dependent retinal atrophy.

Geographic atrophy (GA), the advanced form of dry age-related macular degeneration (AMD), is characterized by progressive loss of retinal pigment epithelium cells and photoreceptors in the setting of characteristic extracellular deposits and remains a serious unmet medical need. While genetic predisposition to AMD is dominated by polymorphisms in complement genes, it remains unclear how complement activation contributes to retinal atrophy. Here we demonstrate that complement is activated on photoreceptor outer segments (POS) in the retina peripheral to atrophic lesions associated with GA. When exposed to human serum following outer blood-retinal barrier breakdown, POS act as potent activators of the classical and alternative complement pathway. In mouse models of retinal degeneration, classical and alternative pathway complement activation on photoreceptors contributed to the loss of photoreceptor function. This was dependent on C5a-mediated recruitment of peripheral blood monocytes but independent of resident microglia. Genetic or pharmacologic inhibition of both classical and alternative complement C3 and C5 convertases was required to reduce progressive degeneration of photoreceptor rods and cones. Our study implicates systemic classical and alternative complement proteins and peripheral blood monocytes as critical effectors of localized retinal degeneration with potential relevance for the contribution of complement activation to GA.

Interactions of physical activity and body mass index with age at menarche: A school-based sample of Chinese female adolescents.

BACKGROUND: To test the independent and interactive associations of physical activity (PA) and body mass index (BMI) with age at menarche among Chinese adolescence. METHODS: We conducted a cross-sectional survey in two elementary schools in September 2015 in Shaoxing city, Zhejiang province, China. We used self-administered questionnaires to collect the information of the participators. Analyses were performed with logistic regression models, and the relative excess risk due to interaction (RERI), the attributable proportion because of the interaction (AP), and the synergy index (S) were used to evaluate the biological interaction. RESULTS: A total of 1530 middle school students from grade 1-3 were selected for this study, and we collected 1505 (98.366%) valid questionnaires for the last analyses. The mean age of menarche is 11.603 (SD=0.447)years. Students with overweight/obesity and without SMPA had the highest risks of early age of menarche (OR=3.507, 95%CI: 1.929-6.376) compared with women both with a normal BMI and with SMPA, and the RERI was 1.846 (95% CI: 0.415-4.107), meaning that there was positive interaction on the additive scale. CONCLUSION: Insufficient PA can significantly modify the association between obesity and early menarche risk. Therefore, the government and society should pay more attention to the promotion of PA throughout childhood and adolescence.

IL-33 amplifies an innate immune response in the degenerating retina.

Age-related macular degeneration (AMD), a leading cause of vision impairment in the ageing population, is characterized by irreversible loss of retinal pigment epithelial (RPE) cells and photoreceptors and can be associated with choroidal neovascularization. Mononuclear phagocytes are often present in AMD lesions, but the processes that direct myeloid cell recruitment remain unclear. Here, we identify IL-33 as a key regulator of inflammation and photoreceptor degeneration after retina stress or injury. IL-33(+) Müller cells were more abundant and IL-33 cytokine was elevated in advanced AMD cases compared with age-matched controls with no AMD. In rodents, retina stress resulted in release of bioactive IL-33 that in turn increased inflammatory chemokine and cytokine expression in activated Müller cells. Deletion of ST2, the IL-33 receptor α chain, or treatment with a soluble IL-33 decoy receptor significantly reduced release of inflammatory mediators from Müller cells, inhibited accumulation of mononuclear phagocytes in the outer retina, and protected photoreceptor rods and cones after a retina insult. This study demonstrates a central role for IL-33 in regulating mononuclear phagocyte recruitment to the photoreceptor layer and positions IL-33 signaling as a potential therapeutic target in macular degenerative diseases.

Micro-Expression Recognition Using Color Spaces.

Micro-expressions are brief involuntary facial expressions that reveal genuine emotions and, thus, help detect lies. Because of their many promising applications, they have attracted the attention of researchers from various fields. Recent research reveals that two perceptual color spaces (CIELab and CIELuv) provide useful information for expression recognition. This paper is an extended version of our International Conference on Pattern Recognition paper, in which we propose a novel color space model, tensor independent color space (TICS), to help recognize micro-expressions. In this paper, we further show that CIELab and CIELuv are also helpful in recognizing micro-expressions, and we indicate why these three color spaces achieve better performance. A micro-expression color video clip is treated as a fourth-order tensor, i.e., a four-dimension array. The first two dimensions are the spatial information, the third is the temporal information, and the fourth is the color information. We transform the fourth dimension from RGB into TICS, in which the color components are as independent as possible. The combination of dynamic texture and independent color components achieves a higher accuracy than does that of RGB. In addition, we define a set of regions of interests (ROIs) based on the facial action coding system and calculated the dynamic texture histograms for each ROI. Experiments are conducted on two micro-expression databases, CASME and CASME 2, and the results show that the performances for TICS, CIELab, and CIELuv are better than those for RGB or gray.

Quantitative histological assessment of xenobiotic-induced liver enzyme induction and pituitary-thyroid axis stimulation in rats using whole-slide automated image analysis.

Preclinical evaluation of a new compound, RO2910, identified a hypertrophic response in liver, thyroid gland, and pituitary gland (pars distalis). We aimed to develop and validate automated image analysis methods to quantify and refine the interpretation of semi-quantitative histology. Wistar-Han rats were administered RO2910 for 14 days. Liver, thyroid, and pituitary gland tissues were processed for routine histology and immunolabeled with anti-thyroid stimulating hormone (TSH) antibody (pituitary) and anti-topoisomerase II antibody (thyroid). Glass slides were scanned, image analysis methods were developed and applied to whole-slide images, and numerical results were compared with histopathology, circulating hormone levels, and liver enzyme mRNA expression for validation. Quantitative analysis of slides had strong individual correlation with semi-quantitative histological evaluation of all tissues studied. Hepatocellular hypertrophy quantification also correlated strongly with liver enzyme mRNA expression. In the pars distalis, measurement of TSH weak-staining areas correlated with both hypertrophy scores and circulating TSH levels. Whole-slide image analysis enabled automated quantification of semi-quantitative histopathology findings and a more refined interpretation of these data. The analysis also enabled a direct correlation with non-histological parameters using straightforward statistical analysis to provide a more refined dose- and sex-response relationship and integration among affected parameters. These findings demonstrate the utility of our image analysis to support preclinical safety evaluations.

Characterization of xenobiotic-induced hepatocellular enzyme induction in rats: anticipated thyroid effects and unique pituitary gland findings.

During routine safety evaluation of RO2910, a non-nucleoside reverse transcriptase inhibitor for HIV infection, histopathology findings concurrent with robust hepatocellular induction occurred in multiple organs, including a unique, albeit related, finding in the pituitary gland. For fourteen days, male and female rats were administered, by oral gavage vehicle, 100, 300, or 1000 mg/kg/day of RO2910. Treated groups had elevated serum thyroid-stimulating hormone and decreased total thyroxine, and hypertrophy in the liver, thyroid gland, and pituitary pars distalis. These were considered consequences of hepatocellular induction and often were dose dependent and more pronounced in males than in females. Hepatocellular centrilobular hypertrophy corresponded with increased expression of cytochrome P450s 2B1/2, 3A1, and 3A2 and UGT 2B1. Bilateral thyroid follicular cell hypertrophy occurred concurrent to increased mitotic activity and sometimes colloid depletion, which were attributed to changes in thyroid hormone levels. Males had hypertrophy of thyroid-stimulating hormone-producing cells (thyrotrophs) in the pituitary pars distalis. All findings were consistent with the well-established adaptive physiologic response of rodents to xenobiotic-induced hepatocellular microsomal enzyme induction. Although the effects on the pituitary gland following hepatic enzyme induction-mediated hypothyroidism have not been reported previously, other models of stress and thyroid depletion leading to pituitary stimulation support such a shared pathogenesis.

Metal contamination in sediments of the western Bohai Bay and adjacent estuaries, China.

Twelve sediment cores were collected in July 2007 in open waters of western Bohai Bay, the Port of Tianjin, and the adjacent estuaries of the Haihe and Yongding Rivers. While overall concentrations of trace metals at incremental depths in these cores met the Marine Sediment Quality (GB18668-2002) criteria of China, the magnitude of both metal enrichment factors (EF) and geoaccumulation indices (I(geo)) suggested that pollution with Ag, Cd, Cr, Cu and Zn was occurring in the estuaries and Port. Risk analysis also suggested that Ag and Ni concentrations were sufficiently elevated as to cause adverse biological effects in the study area. Although metal concentrations in western Bohai Bay were of less concern, a positive relationship between EF values and excess (210)Pb activity for several metals suggested that their concentrations were increasing over time.

Multiplanar computed tomographic analysis of the frontal recess region in Chinese subjects without frontal sinus disease symptoms.

The purposes of this study were to determine the prevalence of frontal-ethmoidal cells and to evaluate variation in the superior attachment of the uncinate process in Chinese subjects. 202 normal Chinese subjects (404 sides) underwent spiral computed tomography and multiplanar reconstruction images were evaluated. Agger nasi cells showed a prevalence of 94.1%. Of all the frontal cells identified in 159 sides (39.6%) of frontal recesses, the prevalence of type I, type II and type III cells was 24.4, 7.0 and 8.2%, respectively. Suprabullar, frontal bullar and interfrontal septal cells were identified in 148 sides (36.6%), 36 sides (9.0%) and 25 subjects (12.4%), respectively. 244 uncinate processes (60.4%) had 1 superior attachment and the remainder (39.6%) had 2 superior attachments. The prevalence of terminal recesses was 89.1%. Our results characterized normal frontal recess pneumatization patterns in Chinese subjects. That, together with variation in the superior attachment of the uncinate process, emphasized the role of agger nasi cells and the uncinate process in endoscopic frontal sinus surgery.