DOA Estimation on One-Bit Quantization Observations through Noise-Boosted Multiple Signal Classification.

Due to the low-complexity implementation, direction-of-arrival (DOA) estimation-based one-bit quantized data are of interest, but also, signal processing struggles to obtain the demanded estimation accuracy. In this study, we injected a number of noise components into the receiving data before the uniform linear array (ULA) composed of one-bit quantizers. Then, based on this designed noise-boosted quantizer unit (NBQU), we propose an efficient one-bit multiple signal classification (MUSIC) method for estimating the DOA. Benefiting from the injected noise, the numerical results show that the proposed NBQU-based MUSIC method outperforms existing one-bit MUSIC methods in terms of estimation accuracy and resolution. Furthermore, with the optimal root mean square (RMS) of the injected noise, the estimation accuracy of the proposed method for estimating DOA can approach that of the MUSIC method based on the complete analog data.

Discovery and validation of combined biomarkers for the diagnosis of esophageal intraepithelial neoplasia and esophageal squamous cell carcinoma.

Early diagnosis and intervention of esophageal squamous cell carcinoma (ESCC) can improve the prognosis. The purpose of this study was to identify biomarkers for ESCC and esophageal precancerous lesions (intraepithelial neoplasia, IEN). Based on the proteomic and genomic data of esophageal tissue including previously reported data, up-regulated proteins with copy number amplification in esophageal cancer were screened as candidate biomarkers. Five proteins, including KDM2A, RAD9A, ECT2, CYHR1 and TONSL, were confirmed by immunohistochemistry on ESCC and normal esophagus (NE). Then, we investigated the expression of 5 proteins in 236 participants (60 NEs, 93 IENs and 83 ESCCs) which were randomly divided into training set and test set. When distinguishing ESCC from NE, the area under curve (AUC) of the multiprotein model was 0.940 in the training set, while the lowest AUC of a protein was 0.735. In the test set, the results were similar. When distinguishing ESCC from IEN or distinguishing IEN from NE, the diagnostic efficiency of the multi-protein models were also improved compared with that of single protein. Our findings suggest that combined detection of KDM2A, RAD9A, ECT2, CYHR1 and TONSL can be used as potential biomarkers for the early diagnosis of ESCC and precancerous lesion development prediction. SIGNIFICANCE: Candidate biomarkers including KDM2A, RAD9A, ECT2, CYHR1 and TONSL screened by integrating genomic and proteomic data from the esophagus can be used as potential biomarkers for the early diagnosis of esophageal squamous cell carcinoma and precancerous lesion development prediction.

The Mechanism of Catalpol to Improve Oxidative Damage of Dermal Fibroblasts Based on Nrf2/HO-1 Signaling Pathway.

Objective: Catalpol, as a natural medicine small-molecule drug, has been proven to have anti-inflammatory and antioxidant pharmacological effects. Methods: The effect of catalpol on oxidative damage of mouse epidermal fibroblast L929 model and its mechanism were investigated by using hydrogen peroxide model, CCK8 method, flow cytometry, and Western blot. Results: The effect of catalpol on Nrf2/HO-1 signaling pathway was further studied to improve oxidative stress in cell models. The results showed that catalpol had no cytotoxicity to L929 cells, and inhibited the apoptosis of L929 cells after oxidative damage in a concentration-dependent manner, thus playing a role in cell protection. The oxidative damage of cells was inhibited by up-regulating the expression of the signature protein of Nrf2/HO-1 signaling pathway and inhibiting the interstitial formation of cells. Conclusion: This study is a preliminary study on the protective function of catalpol against oxidation and apoptosis in dermal fibroblasts, which can provide a theoretical basis and drug guidance for promoting skin wound healing in the later stage.

A druggable cascade links methionine metabolism to epigenomic reprogramming in squamous cell carcinoma.

Upper aerodigestive squamous cell carcinoma (UASCC) is a common and aggressive malignancy with few effective therapeutic options. Here, we investigate amino acid metabolism in this cancer, surprisingly noting that UASCC exhibits the highest methionine level across all human cancers, driven by its transporter LAT1. We show that LAT1 is also expressed at the highest level in UASCC, transcriptionally activated by UASCC-specific promoter and enhancers, which are directly coregulated by SCC master regulators TP63/KLF5/SREBF1. Unexpectedly, unbiased bioinformatic screen identifies EZH2 as the most significant target downstream of the LAT1-methionine pathway, directly linking methionine metabolism to epigenomic reprogramming. Importantly, this cascade is indispensable for the survival and proliferation of UASCC patient-derived tumor organoids. In addition, LAT1 expression is closely associated with cellular sensitivity to inhibition of the LAT1-methionine-EZH2 axis. Notably, this unique LAT1-methionine-EZH2 cascade can be targeted effectively by either pharmacological approaches or dietary intervention in vivo. In summary, this work maps a unique mechanistic cross talk between epigenomic reprogramming with methionine metabolism, establishes its biological significance in the biology of UASCC, and identifies a unique tumor-specific vulnerability which can be exploited both pharmacologically and dietarily.

Developmental toxicity assay of xanthatin in zebrafish embryos.

Xanthatin (XAN), a xanthanolide sesquiterpene lactone, isolated from Chinese herb, Xanthium strumarium L, has various pharmacological activities, such as antitumor activity and anti-inflammatory. However, little is known about its potential toxicity and the mechanism. Here, zebrafish model was used to study the developmental toxicity in vivo. Our results indicated that xanthatin increased the mortality and led to the morphological abnormalities including pericardial edema, yolk sac edema, curved body shape and hatching delay. Furthermore, xanthatin damaged the normal structure and/or function of heart, liver, immune and nervous system. ROS elevation and much more apoptosis cells were observed after xanthatin exposure. Gene expression results showed that oxidative stress-related genes nrf2 was inhibited, while oxidative stress-related genes (keap1 and nqo1) and apoptotic genes (caspase3, caspase9 and p53) were increased after xanthatin exposure. Mitophagy related genes pink1 and parkin, and wnt pathway (ß-catenin, wnt8a and wnt11) were significantly increased after xanthatin exposure. Taken together, our finding indicated that xanthatin induced developmental toxicity, and the ROS elevation, apoptosis activation, dysregulation of mitophagy and wnt pathways were involved in the toxicity caused by xanthatin.

The Chinese keratoconus (CKC) cohort study.

The Chinese keratoconus (CKC) cohort study is a population-based longitudinal prospective cohort study in the Chinese population involving a clinical database and biobanks. This ongoing study focuses on the prevention of KC progression and is the first to involve the effect of gene‒environment interactions on KC progression. The CKC cohort is hospital-based and dynamic and was established in Zhengzhou, China; KC patients (n = 1114) from a large geographical area were enrolled from January 2019 to June 2023, with a mean age of 22.23 years (6‒57 years). Demographic details, socioeconomic characteristics, lifestyle, disease history, surgical history, family history, and visual and social function data are being collected using questionnaires. General physical examination, eye examination, biological specimen collection, and first-degree relative data were collected and analyzed in the present study. The primary focus of the present study was placed on gene, environment and the effect of gene‒environment interactions on KC progression. The follow-up of the CKC cohort study is expected to include data collection at 3 months, 6 months, and 1 year after the initial examination and then at the annual follow-up examinations. The first follow-up of the CKC cohort study was recorded. A total of 918 patients completed the follow-up by June 1, 2023, with a response rate of 82.40%. Aside from the younger age of patients who were followed up, no significant differences were found between patients who were followed up and patients who were not.

Splicing factor hnRNPA1 regulates alternative splicing of LOXL2 to enhance the production of LOXL2Δ13.

Lysyl oxidase-like 2 (LOXL2) is a member of the lysyl oxidase family and has the ability to catalyze the cross-linking of extracellular matrix collagen and elastin. High expression of LOXL2 is related to tumor cell proliferation, invasion, and metastasis. LOXL2 contains 14 exons. Previous studies have found that LOXL2 has abnormal alternative splicing and exon skipping in a variety of tissues and cells, resulting in a new alternatively spliced isoform denoted LOXL2Δ13. LOXL2Δ13 lacks LOXL2WT exon 13, but its encoded protein has greater ability to induce tumor cell proliferation, invasion, and metastasis. However, the molecular events that produce LOXL2Δ13 are still unclear. In this study, we found that overexpression of the splicing factor hnRNPA1 in cells can regulate the alternative splicing of LOXL2 and increase the expression of LOXL2Δ13. The exonic splicing silencer exists at the 3' splice site and 5' splice site of LOXL2 exon 13. HnRNPA1 can bind to the exonic splicing silencer and inhibit the inclusion of exon 13. The RRM domain of hnRNPA1 and phosphorylation of hnRNPA1 at S91 and S95 are important for the regulation of LOXL2 alternative splicing. These results show that hnRNPA1 is a splicing factor that enhances the production of LOXL2Δ13.

Association of Novel Loci With Keratoconus Susceptibility in a Chinese Genome-Wide Association Study.

Purpose: Keratoconus (KC) is a progressive corneal disease that can lead to corneal blindness if not properly managed. The purpose of this study was to identify genetic associations with KC in China and to investigate whether these genetic variants are associated with corneal thickness and corneal curvature in KC cases. Methods: A genome-wide association study was conducted on 853 patients with KC and 6248 controls. The KC cases were genotyped with the Illumina Infinium Human Asian Screening Array BeadChip, and the controls were genotyped with the Illumina Infinium Human Global Screening Array BeadChip. Genetic associations with KC, as well as correlations between the positive variants and corneal parameters including central corneal thickness (CCT) and mean keratometry (Km), were compared using PLINK version 1.90. Results: Our present study identified four single-nucleotide polymorphisms (SNPs) within four risk loci (PTGER3: rs2300163, EYA1: rs1077435, ASS1: rs141365191, and CHTF8: rs3743680) associated with KC in Chinese patients that reached genome-wide significance. Among the identified SNPs with P < 1.00 × 10-4, seven SNPs (FOSL2-PLB1: rs12622211, RXRA-COL5A1: rs3118515, rs3132306, rs1536482, rs3118520, KAT6B: rs192187772, RAP2A-IPO5: rs41361245) were observed to be associated with CCT, and one SNP (USP13: rs6767552) was found to be associated with Km. Conclusions: In the first genome-wide association study of KC with a relatively large study population in China, we identified four SNPs in four risk loci associated with the disease. The findings enriched the understanding of genetic susceptibility to KC and provided new insights into the genetic etiology of the disease.

Construction and validation of classification models for predicting the response to concurrent chemo-radiotherapy of patients with esophageal squamous cell carcinoma based on multi-omics data.

BACKGROUND: Concurrent chemo-radiotherapy (CCRT) is the preferred non-surgical treatment for patients with locally advanced esophageal squamous cell carcinoma (ESCC). Unfortunately, some patients respond poorly, which leads to inappropriate or excessive treatment and affects patient survival. To accurately predict the response of ESCC patients to CCRT, we developed classification models based on the clinical, serum proteomic and radiomic data. METHODS: A total of 138 ESCC patients receiving CCRT were enrolled in this study and randomly split into a training cohort (n = 92) and a test cohort (n = 46). All patients were classified into either complete response (CR) or incomplete response (non-CR) groups according to RECIST1.1. Radiomic features were extracted by 3Dslicer. Serum proteomic data was obtained by Olink proximity extension assay. The logistic regression model with elastic-net penalty and the R-package "rms" v6.2-0 were applied to construct classification and nomogram models, respectively. The area under the receiver operating characteristic curves (AUC) was used to evaluate the predictive performance of the models. RESULTS: Seven classification models based on multi-omics data were constructed, of which Model-COR, which integrates five clinical, five serum proteomic, and seven radiomic features, achieved the best predictive performance on the test cohort (AUC = 0.8357, 95 % CI: 0.7158-0.9556). Meanwhile, patients predicted to be CR by Model-COR showed significantly longer overall survival than those predicted to be non-CR in both cohorts (Log-rank P = 0.0014 and 0.027, respectively). Furthermore, two nomogram models based on multi-omics data also performed well in predicting response to CCRT (AUC = 0.8398 and 0.8483, respectively). CONCLUSION: We developed and validated a multi-omics based classification model and two nomogram models for predicting the response of ESCC patients to CCRT, which achieved the best prediction performance by integrating clinical, serum Olink proteomic, and radiomic data. These models could be useful for personalized treatment decisions and more precise clinical radiotherapy and chemotherapy for ESCC patients.

Long non-coding RNA-NONMMMUT004552.2 regulates the unloading-induced bone loss through the miRNA-15b-5p/Syne1 in mice.

Exercise-induced mechanical loading can increase bone strength whilst mechanical unloading enhances bone-loss. Here, we investigated the role of lncRNA NONMMUT004552.2 in unloading-induced bone-loss. Knockout of lncRNA NONMMUT004552.2 in hindlimb-unloaded mice caused an increase in the bone formation and osteoblast activity. The silencing of lncRNA NONMMUT004552.2 also decreased the osteoblast apoptosis and expression of Bax and cleaved caspase-3, increased Bcl-2 protein expression in MC3T3-E1 cells. Mechanistic investigations demonstrated that NONMMUT004552.2 functions as a competing endogenous RNA (ceRNA) to facilitate the protein expression of spectrin repeat containing, nuclear envelope 1 (Syne1) by competitively binding miR-15b-5p and subsequently inhibits the osteoblast differentiation and bone formation in the microgravity unloading environment. These data highlight the importance of the lncRNA NONMMUT004552.2/miR-15b-5p/Syne1 axis for the treatment of osteoporosis.

KAT8/SIRT7-mediated Fascin-K41 acetylation/deacetylation regulates tumor metastasis in esophageal squamous cell carcinoma.

Fascin actin-bundling protein 1 (Fascin) is highly expressed in a variety of cancers, including esophageal squamous cell carcinoma (ESCC), working as an important oncogenic protein and promoting the migration and invasion of cancer cells by bundling F-actin to facilitate the formation of filopodia and invadopodia. However, it is not clear how exactly the function of Fascin is regulated by acetylation in cancer cells. Here, in ESCC cells, the histone acetyltransferase KAT8 catalyzed Fascin lysine 41 (K41) acetylation, to inhibit Fascin-mediated F-actin bundling and the formation of filopodia and invadopodia. Furthermore, NAD-dependent protein deacetylase sirtuin (SIRT) 7-mediated deacetylation of Fascin-K41 enhances the formation of filopodia and invadopodia, which promotes the migration and invasion of ESCC cells. Clinically, the analysis of cancer and adjacent tissue samples from patients with ESCC showed that Fascin-K41 acetylation was lower in the cancer tissue of patients with lymph node metastasis than in that of patients without lymph node metastasis, and low levels of Fascin-K41 acetylation were associated with a poorer prognosis in patients with ESCC. Importantly, K41 acetylation significantly blocked NP-G2-044, one of the Fascin inhibitors currently being clinically evaluated, suggesting that NP-G2-044 may be more suitable for patients with low levels of Fascin-K41 acetylation, but not suitable for patients with high levels of Fascin-K41 acetylation. © 2024 The Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Fascin - F-actin interaction studied by molecular dynamics simulation and protein network analysis.

Actin bundles are an important component of cellular cytoskeleton and participate in the movement of cells. The formation of actin bundles requires the participation of many actin binding proteins (ABPs). Fascin is a member of ABPs, which plays a key role in bundling filamentous actin (F-actin) to bundles. However, the detailed interactions between fascin and F-actin are unclear. In this study, we construct an atomic-level structure of fascin - F-actin complex based on a rather poor cryo-EM data with resolution of 20 nm. We first optimized the geometries of the complex by molecular dynamics (MD) simulation and analyzed the binding site and pose of fascin which bundles two F-actin chains. Next, binding free energy of fascin was calculated by MM/GBSA method. Finally, protein structure network analysis (PSNs) was performed to analyze the key residues for fascin binding. Our results show that residues of K22, E27, E29, K41, K43, R110, R149, K358, R408 and K471 on fascin are important for its bundling, which are in good agreement with the experimental data. On the other hand, the consistent results indicate that the atomic-level model of fascin - F-actin complex is reliable. In short, this model can be used to understand the detailed interactions between fascin and F-actin, and to develop novel potential drugs targeting fascin.Communicated by Ramaswamy H. Sarma.

CCL2 Potentiates Inflammation Pain and Related Anxiety-Like Behavior Through NMDA Signaling in Anterior Cingulate Cortex.

Previous studies have shown that the C-C motif chemokine ligand 2 (CCL2) is widely expressed in the nervous system and involved in regulating the development of chronic pain and related anxiety-like behaviors, but its precise mechanism is still unclear. This paper provides an in-depth examination of the involvement of CCL2-CCR2 signaling in the anterior cingulate cortex (ACC) in intraplantar injection of complete Freund's adjuvant (CFA) leading to inflammatory pain and its concomitant anxiety-like behaviors by modulation of glutamatergic N-methyl-D-aspartate receptor (NMDAR). Our findings suggest that local bilateral injection of CCR2 antagonist in the ACC inhibits CFA-induced inflammatory pain and anxiety-like behavior. Meanwhile, the expression of CCR2 and CCL2 was significantly increased in ACC after 14 days of intraplantar injection of CFA, and CCR2 was mainly expressed in excitatory neurons. Whole-cell patch-clamp recordings showed that the CCR2 inhibitor RS504393 reduced the frequency of miniature excitatory postsynaptic currents (mEPSC) in ACC, and CCL2 was involved in the regulation of NMDAR-induced current in ACC neurons in the pathological state. In addition, local injection of the NR2B inhibitor of NMDAR subunits, Ro 25-6981, attenuated the effects of CCL2-induced hyperalgesia and anxiety-like behavior in the ACC. In summary, CCL2 acts on CCR2 in ACC excitatory neurons and participates in the regulation of CFA-induced pain and related anxiety-like behaviors through upregulation of NR2B. CCR2 in the ACC neuron may be a potential target for the treatment of chronic inflammatory pain and pain-related anxiety.

Machine learning analysis with the comprehensive index of corneal tomographic and biomechanical parameters in detecting pediatric subclinical keratoconus.

Background: Keratoconus (KC) occurs at puberty but diagnosis is focused on adults. The early diagnosis of pediatric KC can prevent its progression and improve the quality of life of patients. This study aimed to evaluate the ability of corneal tomographic and biomechanical variables through machine learning analysis to detect subclinical keratoconus (SKC) in a pediatric population. Methods: Fifty-two KC, 52 SKC, and 52 control pediatric eyes matched by age and gender were recruited in a case-control study. The corneal tomographic and biomechanical parameters were measured by professionals. A linear mixed-effects test was used to compare the differences among the three groups and a least significant difference analysis was used to conduct pairwise comparisons. The receiver operating characteristic (ROC) curve and the Delong test were used to evaluate diagnostic ability. Variables were used in a multivariate logistic regression in the machine learning analysis, using a stepwise variable selection to decrease overfitting, and comprehensive indices for detecting pediatric SKC eyes were produced in each step. Results: PE, BAD-D, and TBI had the highest area under the curve (AUC) values in identifying pediatric KC eyes, and the corresponding cutoff values were 12 µm, 2.48, and 0.6, respectively. For discriminating SKC eyes, the highest AUC (95% CI) was found in SP A1 with a value of 0.84 (0.765, 0.915), and BAD-D was the best parameter among the corneal tomographic parameters with an AUC (95% CI) value of 0.817 (0.729, 0.886). Three models were generated in the machine learning analysis, and Model 3 (y = 0.400*PE + 1.982* DA ratio max [2 mm]-0.072 * SP A1-3.245) had the highest AUC (95% CI) value, with 90.4% sensitivity and 76.9% specificity, and the cutoff value providing the best Youden index was 0.19. Conclusion: The criteria of parameters for diagnosing pediatric KC and SKC eyes were inconsistent with the adult population. Combined corneal tomographic and biomechanical parameters could enhance the early diagnosis of young patients and improve the inadequate representation of pediatric KC research.

Elevated serum creatinine levels and risk of cognitive impairment in older adults with diabetes: a NHANES study from 2011-2014.

Background: The brain and kidney have similar microvascular structure, which makes them susceptible to certain common pathophysiological processes. In this study, we examined several indicators of kidney injury/function associated with cognitive function in older diabetic patients in the hope of finding effective markers for detecting cognitive impairment (CI). Methods: A total of 2209 older participants (aged ≥60 years) from the 2011-2014 National Health and Nutrition Examination Survey (NHANES) were analyzed for the association between diabetes and CI using a multiple linear regression analysis model. Using the same approach, we also analyzed the relationship between indicators of kidney injury/function and cognitive function (Animal Fluency Test, Digit Symbol Substitution Test) in the diabetic population. Results: Diabetes was associated with CI. In age-adjusted model, older diabetics performed significantly poorer on tests of cognitive function compared to normoglycaemic individuals (1.145 points lower on the Animal Fluency Test (P = 0.005) and 7.868 points reduced on the Digit Symbol Substitution Test (P < 0.001)). In diabetics, we found elevated serum creatinine (SCr) (especially at SCr≥300uM) was associated with lower scores on cognitive function tests after strict adjustment for potential influences on cognitive function. While, albumin/creatinine ratio (ACR) was only associated with Digit Symbol Substitution score (DSS) not Animal Fluency score (AFS), and estimated glomerular filtration rate (eGFR) was only associated with CI (AFS and DSS) at the end-stage renal disease. Conclusion: SCr, as a sensitive indicator of kidney injury, was significantly associated with CI and can potentially be used as an effective marker for screening CI in older diabetics.

Mitochondrial DNA heteroplasmy analysis in keratoconus patients from China.

Background: Mitochondrial DNA (mtDNA) variants have been implicated in keratoconus (KC). The present study aimed to characterize the mtDNA heteroplasmy profile in KC and explore the association of mitochondrial heteroplasmic levels with KC. Methods: Mitochondrial sequencing of peripheral blood samples and corneal tomography were conducted in 300 KC cases and 300 matched controls. The number of heteroplasmic and homoplasmic variants was calculated across the mitochondrial genome. Spearman's correlation was used to analyze the correlation between the number of heteroplasmic variants and age. The association of mtDNA heteroplasmic level with KC was analyzed by logistic regression analysis. Moreover, the relationship between mitochondrial heteroplasmic levels and clinical parameters was determined by linear regression analysis. Results: The distribution of mtDNA heteroplasmic variants showed the highest number of heteroplasmic variants in the non-coding region, while the COX3 gene exhibited the highest number in protein-coding genes. Comparisons of the number of heteroplasmic and homoplasmic non-synonymous variants in protein-coding genes revealed no significant differences between KC cases and controls (all p > 0.05). In addition, the number of heteroplasmic variants was positively associated with age in all subjects (r = 0.085, p = 0.037). The logistic regression analyses indicated that the heteroplasmic levels of m.16180_16181delAA was associated with KC (p < 0.005). Linear regression analyses demonstrated that the heteroplasmic levels of m.16180_16181delAA and m.302A>C were not correlated with thinnest corneal thickness (TCT), steep keratometry (Ks), and flat keratometry (Kf) (all p > 0.05) in KC cases and controls separately. Conclusion: The current study characterized the mtDNA heteroplasmy profile in KC, and revealed that the heteroplasmic levels of m.16180_16181delAA were associated with KC.

Improved photoacoustic images via wavefront shaping modulation based on the scattering structure: retraction.

The referenced article [Opt. Express30, 36489 (2022)10.1364/OE.470330] has been retracted by the authors.

A high body mass index strengthens the association between the time of eye rubbing and keratoconus in a Chinese population: a case control study.

BACKGROUND: Although body mass index (BMI) and eye rubbing are linked to an increased risk of keratoconus (KC), the interactive effect of eye rubbing and BMI on KC is largely unknown. This study aimed to evaluate the independent and interactive effects of BMI and eye rubbing on KC and to further explore the role of environmental factors on the occurrence of KC. METHODS: A total of 621 individuals (291 KC patients and 330 control individuals) were enrolled in this hospital­based study on KC patients in Central China after individuals missing BMI data were excluded. BMI was calculated as weight in kilograms divided by the square of height in meters. Data on eye rubbing was recorded through face-to-face interviews. Generalized linear regression models were used to analyze associations among BMI, eye rubbing and KC. Interaction plots were used to describe the interactive effects of BMI and eye rubbing on KC. RESULTS: The ß and 95% confidence interval (CI) were 0.923 (0.112, 1.733) (p = 0.026) and 3.356 (1.953, 4.759) (p < 0.001), respectively, for the effect of each 10 kg/m2 increase in BMI and each 1 min increase in eye rubbing on KC. The interaction of BMI and eye rubbing were positively correlated with KC (p < 0.001). CONCLUSION: These findings suggested that a high BMI aggravated the negative effect of eye rubbing on KC, implying that individuals with a high BMI may be more susceptible to exposure to eye rubbing, which is related to an increased risk of KC.

Masked Graph Neural Networks for Unsupervised Anomaly Detection in Multivariate Time Series.

Anomaly detection has been widely used in grid operation and maintenance, machine fault detection, and so on. In these applications, the multivariate time-series data from multiple sensors with latent relationships are always high-dimensional, which makes multivariate time-series anomaly detection particularly challenging. In existing unsupervised anomaly detection methods for multivariate time series, it is difficult to capture the complex associations among multiple sensors. Graph neural networks (GNNs) can model complex relations in the form of a graph, but the observed time-series data from multiple sensors lack explicit graph structures. GNNs cannot automatically learn the complex correlations in the multivariate time-series data or make good use of the latent relationships among time-series data. In this paper, we propose a new method-masked graph neural networks for unsupervised anomaly detection (MGUAD). MGUAD can learn the structure of the unobserved causality among sensors to detect anomalies. To robustly learn the temporal context from adjacent time points of time-series data from the same sensor, MGUAD randomly masks some points of the time-series data from the sensor and reconstructs the masked time points. Similarly, to robustly learn the graph-level context from adjacent nodes or edges in the relation graph of multivariate time series, MGUAD masks some nodes or edges in the graph under the framework of a GNN. Comprehensive experiments are conducted on three public datasets. According to the experimental findings, MGUAD outperforms state-of-the-art anomaly detection methods.