Search

Background: Anoikis is a specialized form of programmed cell death induced by the loss of cell adhesion to the extracellular matrix (ECM). Acquisition of anoikis resistance is a significant marker for cancer cell invasion, metastasis, therapy resistance, and recurrence. Although current research has identified multiple factors that regulate anoikis resistance, the pathological mechanisms of anoikis-mediated tumor microenvironment (TME) in glioblastoma (GBM) remain largely unexplored. Methods: Utilizing single-cell RNA sequencing (scRNA-seq) data and employing non-negative matrix factorization (NMF), we identified and characterized TME cell clusters with distinct anoikis-associated gene signatures. Prognostic and therapeutic response analyses were conducted using TCGA and CGGA datasets to assess the clinical significance of different TME cell clusters. The spatial relationship between BRMS1 + microglia and tumor cells was inferred from spatial transcriptome RNA sequencing (stRNA-seq) data. To simulate the tumor immune microenvironment, co-culture experiments were performed with microglia (HMC3) and GBM cells (U118/U251), and microglia were transfected with a BRMS1 overexpression lentivirus. Western blot or ELISA were used to detect BRMS1, M2 macrophage-specific markers, PI3K/AKT signaling proteins, and apoptosis-related proteins. The proliferation and apoptosis capabilities of tumor cells were evaluated using CCK-8, colony formation, and apoptosis assays, while the invasive and migratory abilities of tumor cells were assessed using Transwell assays. Results: NMF-based analysis successfully identified CD8 + T cell and microglia cell clusters with distinct gene signature characteristics. Trajectory analysis, cell communication, and gene regulatory network analyses collectively indicated that anoikis-mediated TME cell clusters can influence tumor cell development through various mechanisms. Notably, BRMS1 + AP-Mic exhibited an M2 macrophage phenotype and had significant cell communication with malignant cells. Moreover, high expression of BRMS1 + AP-Mic in TCGA and CGGA datasets was associated with poorer survival outcomes, indicating its detrimental impact on immunotherapy. Upregulation of BRMS1 in microglia may lead to M2 macrophage polarization, activate the PI3K/AKT signaling pathway through SPP1/CD44-mediated cell interactions, inhibit tumor cell apoptosis, and promote tumor proliferation and invasion. Conclusion: This pioneering study used NMF-based analysis to reveal the important predictive value of anoikis-regulated TME in GBM for prognosis and immunotherapeutic response. BRMS1 + microglial cells provide a new perspective for a deeper understanding of the immunosuppressive microenvironment of GBM and could serve as a potential therapeutic target in the future. [ABSTRACT FROM AUTHOR]

Lupus nephritis (LN) is a challenging condition with limited diagnostic and treatment options. In this study, we applied 12 distinct machine learning algorithms along with Non-negative Matrix Factorization (NMF) to analyze single-cell datasets from kidney biopsies, aiming to provide a comprehensive profile of LN. Through this analysis, we identified various immune cell populations and their roles in LN progression and constructed 102 machine learning-based immune-related gene (IRG) predictive models. The most effective models demonstrated high predictive accuracy, evidenced by Area Under the Curve (AUC) values, and were further validated in external cohorts. These models highlight six hub IRGs (CD14, CYBB, IFNGR1, IL1B, MSR1, and PLAUR) as key diagnostic markers for LN, showing remarkable diagnostic performance in both renal and peripheral blood cohorts, thus offering a novel approach for noninvasive LN diagnosis. Further clinical correlation analysis revealed that expressions of IFNGR1, PLAUR, and CYBB were negatively correlated with the glomerular filtration rate (GFR), while CYBB also positively correlated with proteinuria and serum creatinine levels, highlighting their roles in LN pathophysiology. Additionally, protein-protein interaction (PPI) analysis revealed significant networks involving hub IRGs, emphasizing the importance of the interleukin family and chemokines in LN pathogenesis. This study highlights the potential of integrating advanced genomic tools and machine learning algorithms to improve diagnosis and personalize management of complex autoimmune diseases like LN. [ABSTRACT FROM AUTHOR]

Pulsed eddy current thermography can detect surface or subsurface defects in steel, but in the process of combining deep learning, it is expensive and inefficient to build a complete sample of defects due to the complexity of the actual industrial environment. Consequently, this study proposes a transfer learning method based on Twin-NMF and combines it with the SimAM attention mechanism to enhance the detection accuracy of the target domain task. First, to address the domain differences between the target domain task and the source domain samples, this study introduces a Twin-NMF transfer method. This approach reconstructs the feature space of both the source and target domains using twin non-negative matrix factorization and employs cosine similarity to measure the correlation between the features of these two domains. Secondly, this study integrates a parameter-free SimAM into the neck of the YOLOv8 model to enhance its capabilities in extracting and classifying steel surface defects, as well as to alleviate the precision collapse phenomenon associated with multi-scale defect recognition. The experimental results show that the proposed Twin-NMF model with SimAM improves the detection accuracy of steel surface defects. Taking NEU-DET and GC10-DET as source domains, respectively, in the ECTI dataset, mAP@0.5 reaches 99.3% and 99.2%, and the detection accuracy reaches 98% and 98.5%. [ABSTRACT FROM AUTHOR]

Dimension reduction is one of the most sought-after methodologies to deal with high-dimensional ever-expanding complex datasets. Non-negative matrix factorization (NMF) is one such technique for dimension reduction. Here, a multiple deconstruction multiple reconstruction deep learning model (DN3MF) for NMF targeted towards low rank approximation, has been developed. Non-negative input data has been processed using hierarchical learning to generate part-based sparse and meaningful representation. The novel design of DN3MF ensures the non-negativity requirement of the model. The use of Xavier initialization technique solves the exploding or vanishing gradient problem. The objective function of the model has been designed employing regularization, ensuring the best possible approximation of the input matrix. A novel adaptive learning mechanism has been developed to accomplish the objective of the model. The superior performance of the proposed model has been established by comparing the results obtained by the model with that of six other well-established dimension reduction algorithms on three well-known datasets in terms of preservation of the local structure of data in low rank embedding, and in the context of downstream analyses using classification and clustering. The statistical significance of the results has also been established. The outcome clearly demonstrates DN3MF’s superiority over compared dimension reduction approaches in terms of both statistical and intrinsic property preservation standards. The comparative analysis of all seven dimensionality reduction algorithms including DN3MF with respect to the computational complexity and a pictorial depiction of the convergence analysis for both stages of DN3MF have also been presented. [ABSTRACT FROM AUTHOR]

In this research, a GAmified Mobile Leaning Framework (GAMOLEAF) developed as a new intelligent application designed for mobile devices to ensure learning, assessing, and advancing learners' knowledge in programming complex data structures in Java programming language. GAMOLEAF adopted motivational strategies to solve motivational problems during the COVID-19 pandemic by employing a gamification module, that integrates levels, scores, badges, leaderboard, and feedback. Furthermore, in order to assist learners to find useful and relevant lessons and best solutions for each data structure, GAMOLEAF incorporated personalized recommendations through two intelligent modules: a Lessons Recommendation Module (LRecM) and a problem-solving Solutions Recommender Module (PSSORecM). LrecM aims to provide learners with personalized lessons depending on the ratings collected explicitly from them. Whereas, PSSORecM bases on learners' behaviors and directs them to consult other solutions. Both modules were based on the collaborative filtering method and used Matrix Factorization (MF) applying Singular Value Decomposition (SVD) and Negative Matrix Factorization (NMF) algorithms, respectively. To explore how the integration of personalized recommendations and gamification impact on students motivation and learning achievements in higher education to learning programming complex data structures course using mobile technologies, especially in difficult times like COVID-19, an experiment was carried out to compare the learning achievement and motivation of 90 students divided into three groups (control group, first experimental group, and second experimental group) using three versions of GAMOLEAF respectively: GAMOLEAF-V1 without gamification and without recommendation, GAMOLEAF-V2 integrating gamification only and GAMOLEAF-V3 integrating both gamification and recommendation. The One-way ANOVA (analysis of variance) test and Post hoc Tukey test were employed to analyze the performances of the three groups before and after the learning activity. The results suggested that the students who learned with GAMOLEAF-V3 using gamification and recommendation achieved significantly better learning achievement than those who learned with GAMOLEAF-V2 and GAMOLEAF-V1. From the experimental results, it was found that the gamification applied in GAMOLEAF-V2 and GAMOLEAF-V3 had significantly better effectiveness in improving only students' motivation without improving their learning achievement. Moreover, the analysis result of the learning achievement indicated that the students in the second experimental group showed significantly higher learning achievement using GAMOLEAF-V3 in comparison with those in both the control group and the first experimental group who used GAMOLEAF-V1 and GAMOLEAF-V2 respectively. Such findings indicate that the personalized recommendations offered by the Lessons Recommendation Module (LRecM) and the problem-solving Solutions Recommender Module (PSSORecM) in GAMOLEAF-V3 may be one of the reasons why the learning achievement of students was increased. [ABSTRACT FROM AUTHOR]

Dynamic functional connectivity (DFC) analysis using functional magnetic resonance imaging (fMRI) technology has attracted increasing attention in revealing brain dynamics in recent years. Although the nonnegative matrix factorization (NMF) method was applied to dynamic subgraph analysis to reveal brain dynamics, its application in DFC analysis was largely limited due to its nonnegative constraint on the input data. This study proposed the extended NMF (eNMF) method that allowed the input matrix and decomposed basis matrix to have negative values without altering the NMF algorithm. The eNMF method was applied to DFC analysis of both simulated and real resting fMRI data. The simulated data demonstrated that eNMF successfully decomposed the mixed-sign matrix into one positive matrix and one mixed-sign matrix. In contrast to K-means, eNMF extracted more accurate brain state patterns in all cases and estimated better DFC temporal properties for uneven brain state distribution. The real resting-fMRI data demonstrated that eNMF can provide more temporal measures of DFC and was more sensitive to detect intergroup differences of DFC than K-means. Results of eNMF revealed that the female group possibly showed worse relaxation and produced stronger spontaneous cognitive processes although they tended to spend more time in relaxation state and less time in states relevant to cognitive processes in contrast to the male group. [ABSTRACT FROM AUTHOR]

This study introduces an unsupervised term extraction approach that combines non-negative matrix factorization (NMF) with word embeddings. Inspired by a pioneering semantic NMF method that employs regularization to jointly optimize document–word and word–word matrix factorizations for document clustering, we adapt this strategy for term extraction. Typically, a word–word matrix representing semantic relationships between words is constructed using cosine similarities between word embeddings. However, it has been established that transformer encoder embeddings tend to reside within a narrow cone, leading to consistently high cosine similarities between words. To address this issue, we replace the conventional word–word matrix with a word–seed submatrix, restricting columns to 'domain seeds'—specific words that encapsulate the essential semantic features of the domain. Therefore, we propose a modified NMF framework that jointly factorizes the document–word and word–seed matrices, producing more precise encoding vectors for words, which we utilize to extract high-relevancy topic-related terms. Our modification significantly improves term extraction effectiveness, marking the first implementation of semantically enhanced NMF, designed specifically for the task of term extraction. Comparative experiments demonstrate that our method outperforms both traditional NMF and advanced transformer-based methods such as KeyBERT and BERTopic. To support further research and application, we compile and manually annotate two new datasets, each containing 1000 sentences, from the 'Geography and History' and 'National Heroes' domains. These datasets are useful for both term extraction and document classification tasks. All related code and datasets are freely available. [ABSTRACT FROM AUTHOR]

The link between ferroptosis, a form of cell death mediated by iron and acute kidney injury (AKI) is recently gaining widespread attention. However, the mechanism of the crosstalk between cells in the pathogenesis and progression of acute kidney injury remains unexplored. In our research, we performed a non-negative matrix decomposition (NMF) algorithm on acute kidney injury single-cell RNA sequencing data based specifically focusing in ferroptosis-associated genes. Through a combination with pseudo-time analysis, cell–cell interaction analysis and SCENIC analysis, we discovered that proximal tubular cells, macrophages, and fibroblasts all showed associations with ferroptosis in different pathways and at various time. This involvement influenced cellular functions, enhancing cellular communication and activating multiple transcription factors. In addition, analyzing bulk expression profiles and marker genes of newly defined ferroptosis subtypes of cells, we have identified crucial cell subtypes, including Egr1 + PTC-C1, Jun + PTC-C3, Cxcl2 + Mac-C1 and Egr1 + Fib-C1. All these subtypes which were found in AKI mice kidneys and played significantly distinct roles from those of normal mice. Moreover, we verified the differential expression of Egr1, Jun, and Cxcl2 in the IRI mouse model and acute kidney injury human samples. Finally, our research presented a novel analysis of the crosstalk of proximal tubular cells, macrophages and fibroblasts in acute kidney injury targeting ferroptosis, therefore, contributing to better understanding the acute kidney injury pathogenesis, self-repairment and acute kidney injury-chronic kidney disease (AKI-CKD) progression. [ABSTRACT FROM AUTHOR]

Purpose: This paper aims to compare the mission statements of Indian and Singaporean firms in the healthcare sector, and define the main components of Indian and Singaporean mission statements. Design/methodology/approach: The study was based on a network analytic approach and content analysis. The research was performed on 200 companies (100 Indian companies and 100 Singaporean companies). For each company, we searched for a mission statement published in the company website. Nonnegative Matrix Factorization (NMF) in Python programming language was utilized to obtain the differences in the components of mission statements between Indian and Singaporean firms. Findings: The study results indicate a similarity and variation between Indian and Singaporean mission statements. Both countries are more concerned about patients, service, community, quality, and healthcare in their mission statements, but Indian mission statements emphasize quality, affordable price, and technology more than Singaporean firms. In contrast, Singaporean mission statements tend to highlight innovation and company value. This research will assist strategic managers in identifying the mission statement components and choosing the right strategy for the organization. Originality/value: This study contributes to the literature and ethos theory by identifying and distinguishing the paramount differences between the Indian and Singaporean mission statement components in the healthcare sector. [ABSTRACT FROM AUTHOR]

INTRODUCTION: Previous approaches pursuing in vivo staging of tau pathology in Alzheimer's disease (AD) have typically relied on neuropathologically defined criteria. In using predefined systems, these studies may miss spatial deposition patterns which are informative of disease progression. METHODS: We selected discovery (n = 418) and replication (n = 132) cohorts with flortaucipir imaging. Non‐negative matrix factorization (NMF) was applied to learn tau covariance patterns and develop a tau staging system. Flortaucipir components were also validated by comparison with amyloid burden, gray matter loss, and the expression of AD‐related genes. RESULTS: We found eight flortaucipir covariance patterns which were reproducible and overlapped with relevant gene expression maps. Tau stages were associated with AD severity as indexed by dementia status and neuropsychological performance. Comparisons of flortaucipir uptake with amyloid and atrophy also supported our model of tau progression. DISCUSSION: Data‐driven decomposition of flortaucipir uptake provides a novel framework for tau staging which complements existing systems. Highlights: NMF reveals patterns of tau deposition in AD.Data‐driven staging of flortaucipir tracks AD severity.Learned flortaucipir patterns overlap with AD‐related gene expression. [ABSTRACT FROM AUTHOR]

AIMS/HYPOTHESIS: Type 2 diabetes is highly polygenic and influenced by multiple biological pathways. Rapid expansion in the number of type 2 diabetes loci can be leveraged to identify such pathways. METHODS: We developed a high-throughput pipeline to enable clustering of type 2 diabetes loci based on variant-trait associations. Our pipeline extracted summary statistics from genome-wide association studies (GWAS) for type 2 diabetes and related traits to generate a matrix of 323 variants × 64 trait associations and applied Bayesian non-negative matrix factorisation (bNMF) to identify genetic components of type 2 diabetes. Epigenomic enrichment analysis was performed in 28 cell types and single pancreatic cells. We generated cluster-specific polygenic scores and performed regression analysis in an independent cohort (N=25,419) to assess for clinical relevance. RESULTS: We identified ten clusters of genetic loci, recapturing the five from our prior analysis as well as novel clusters related to beta cell dysfunction, pronounced insulin secretion, and levels of alkaline phosphatase, lipoprotein A and sex hormone-binding globulin. Four clusters related to mechanisms of insulin deficiency, five to insulin resistance and one had an unclear mechanism. The clusters displayed tissue-specific epigenomic enrichment, notably with the two beta cell clusters differentially enriched in functional and stressed pancreatic beta cell states. Additionally, cluster-specific polygenic scores were differentially associated with patient clinical characteristics and outcomes. The pipeline was applied to coronary artery disease and chronic kidney disease, identifying multiple overlapping clusters with type 2 diabetes. CONCLUSIONS/INTERPRETATION: Our approach stratifies type 2 diabetes loci into physiologically interpretable genetic clusters associated with distinct tissues and clinical outcomes. The pipeline allows for efficient updating as additional GWAS become available and can be readily applied to other conditions, facilitating clinical translation of GWAS findings. Software to perform this clustering pipeline is freely available.

Background: Proliferative diabetic retinopathy (PDR), a major cause of blindness, is characterized by complex pathogenesis. This study integrates single-cell RNA sequencing (scRNA-seq), Non-negative Matrix Factorization (NMF), machine learning, and AlphaFold 2 methods to explore the molecular level of PDR. Methods: We analyzed scRNA-seq data from PDR patients and healthy controls to identify distinct cellular subtypes and gene expression patterns. NMF was used to define specific transcriptional programs in PDR. The oxidative stress-related genes (ORGs) identified within Meta-Program 1 were utilized to construct a predictive model using twelve machine learning algorithms. Furthermore, we employed AlphaFold 2 for the prediction of protein structures, complementing this with molecular docking to validate the structural foundation of potential therapeutic targets. We also analyzed protein-protein interaction (PPI) networks and the interplay among key ORGs. Results: Our scRNA-seq analysis revealed five major cell types and 14 subcell types in PDR patients, with significant differences in gene expression compared to those in controls. We identified three key meta-programs underscoring the role of microglia in the pathogenesis of PDR. Three critical ORGs (ALKBH1, PSIP1, and ATP13A2) were identified, with the best-performing predictive model demonstrating high accuracy (AUC of 0.989 in the training cohort and 0.833 in the validation cohort). Moreover, AlphaFold 2 predictions combined with molecular docking revealed that resveratrol has a strong affinity for ALKBH1, indicating its potential as a targeted therapeutic agent. PPI network analysis, revealed a complex network of interactions among the hub ORGs and other genes, suggesting a collective role in PDR pathogenesis. Conclusion: This study provides insights into the cellular and molecular aspects of PDR, identifying potential biomarkers and therapeutic targets using advanced technological approaches. [ABSTRACT FROM AUTHOR]

Copyright of Polish Statistician / Wiadomości Statystyczne is the property of State Treasury - Statistics Poland and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

This paper proposes a dual-channel speech separation method based on spatial estimation via sparse Bayesian inference (SBI) and nonnegative matrix factorization (NMF). The spatial information estimated by traditional compressed sensing (CS) models is insufficient when two microphones receive limited columns of mixed signals. Considering the sparsity of peak values in the cross-correlation spectrum between two received signals, the proposed method builds a new CS model based on cross-correlation spectrum and applies SBI algorithm to solve this model to improve the estimation accuracy of spatial information. Combined the spatial information with the spectral features decomposed by NMF, NMF coefficient matrix masks belonging to individual source are generated for pre-separation. To mitigate retained potential interference components, a post-separation processing stage is designed using an expectation maximization (EM) algorithm based on a Gaussian mixture model (GMM). The estimated spatial information and binary time–frequency masks are used for parameter initialization of the EM algorithm. The experimental results using real-world speech data show that the proposed method can achieve better separation performance compared to various existing methods. [ABSTRACT FROM AUTHOR]

We propose a novel single channel speech enhancement algorithm using iterative constrained Non-negative matrix factorization (NMF) based adaptive Wiener gain for non-stationary noise. In the recent past, NMF-based Wiener filtering methods were used for speech enhancement. The Wiener filter performance depends on the adaptive gain factor value. The adaptive gain factor (α ) value is constant regardless of noise type and signal to noise ratio (SNR), so it will affect speech enhancement performance. To overcome this, the adaptive factor value is calculated using a genetic algorithm (GA). Here, the GA adjusts the adaptive Wiener gain based on noise type and SNR level. The GA-based adaptive Wiener gain minimizes Wiener filter estimation errors and improves speech quality by adjusting the base vector weights of noise and speech. Additionally, we use the iterative constraints NMF (IC-NMF) method for calculating the priors from noisy speech magnitudes. We select the Erlang, Inverse Gamma, Students-t, and Inverse Nakagami distributions for speech priors and Gaussian distributions for noise priors. Noise and speech samples are well correlated with those distributions. This provides accurate estimation of the necessary statistics of these distributions to regularize the NMF criterion. So, we combine an iterative constrained NMF and a genetic algorithm-based adaptive Wiener filtering method for speech enhancement. The proposed method outperforms other benchmark algorithms in terms of source to distortion ratio (SDR), short-time objective intelligibility (STOI), and perceptual evaluation of speech quality (PESQ). [ABSTRACT FROM AUTHOR]

Link prediction aims to predict missing links or eliminate spurious links and anticipate new links by analyzing observed network topological structure information. Non-negative matrix factorization(NMF) is widely used in solving the issue of link prediction due to its good interpretability and scalability. However, most existing NMF-based approaches involve shallow decoder models, which are incapable of capturing complex hierarchical information hidden in networks, and seldom consider random noise. To address these issues, a novel deep autoencoder-like nonnegative matrix factorization method with L 2 , 1 norm for link prediction models(DANMFL) is proposed. Unlike conventional NMF-based approaches, our model contains a decoder component and an encoder component, which capture complex hierarchical information effectively, leading to more accurate prediction results. In addition, we use the L 2 , 1 norm to remove random noise in each layer and the convergence of our model is strictly proven. We conduct extensive experiments on twelve real-world networks and the experimental results show that DANMFL is superior to existing state-of-the-art baseline approaches in terms of prediction accuracy. Codes are available at https://github.com/litongf/DANMFL. [ABSTRACT FROM AUTHOR]

At present, intelligent systems with the function of automatic micro expression recognition are gradually applied in intelligent classrooms, but there is still a problem of low recognition rate. Therefore, based on the dual graph regularization, the research constructs a joint non negative matrix decomposition algorithm model for micro expression recognition, and verifies its effectiveness. The experimental results showed that the research algorithm had the highest performance in both micro and macro expression (micro) databases, at 75.4 %. In algorithm comparison, the recognition rates of algorithm L and algorithm M in different databases were higher than those of group A and B algorithms. Among them, Algorithm 6 in Group A had the highest recognition rate in all three databases, with the highest being 57.4 %; Algorithm 12 in Group B had the highest recognition rate in Zhongke Microexpression 2 and micro and macro expression databases, with 61.1 % and 59.4 %; Algorithm 11 had the highest recognition rate in self generated micro expression databases, with 54.0 %. And algorithm L and algorithm M had a minimum of 58.5 % and a maximum of 75.4 %. In parameter sensitivity analysis, the recognition rates of parameters η, χ, and γ in all databases showed good recognition results within a certain range of values, but they would decrease when they exceeded a specific value. Overall, the algorithm model proposed in the study has high effectiveness in improving the recognition rate of micro expressions, which is significant for the micro expression recognition of students in practical intelligent classrooms.

High dropout rates globally perpetuate educational disparities with various underlying causes. Despite numerous strategies to address this issue, more attention should be given to understanding and addressing student emotions during classes. This lack of focus adversely affects learner engagement and retention rates. While previous studies on online learning have primarily emphasized the effectiveness of technology, infrastructure, cognition, motivation, and economic benefits, there is still a gap in understanding the emotional aspects of distance learning. First, this study addresses this gap by employing thematic modeling and utilizing non-negative matrix factorization (NMF) for emotion recognition through students’ deep learning techniques and facial emotion recognition (FER). Second, statistical analysis of these findings further augments the depth of the study. Finally, the research proposes a mathematical model based on the random walk of emotional state transitions. The findings of this study underscore the importance of considering emotions in distance learning environments and their significant impact on student’s academic performance and satisfaction. By acknowledging and addressing these emotional factors, educators can enhance learner engagement, promote positive emotions, mitigate negative emotions during online learning, and ultimately improve the effectiveness of online courses.

Due to the addition of new underwater vessels and other natural noise contributors, the underwater environment is becoming congested and noisy. Undersea monitoring sonobuoys receive multiple mixed acoustic signals from different vessels that need to be separated and identified in the presence of underwater noise (UWN). It is extremely challenging to separate highly correlated acoustic signals from a noisy mixture without prior knowledge of mixing process and propagation channel. Also, in many cases, the separated signals from the noisy mixture doesn’t accurately describe the correct signal. This study proposes a novel multi-stage method to separate underwater acoustic source signals from noisy mixture with suppression in noise. The first stage applies multivariate blind source separation (BSS) technique known as non-negative matrix factorization (NMF) that extracts the source signals from the noisy signal mixture. In the second stage, Minimum mean square error (MMSE) estimator is used to reduce noise in separated/reconstructed source signals by minimizing the mean square error (MSE) between reconstructed acoustic signal and original clean signal, enhancing signal reconstruction quality. The results of this study indicate the effectiveness of the proposed method in terms of MSE, signal-to-distortion ratio (SDR) and cepstral distance (CD) and compare it with existing techniques. Based on simulation outcomes our proposed method demonstrates superior separation performance by reducing MSE upto 47% and improving SDR of reconstructed acoustic signals upto 28% compared to existing solutions.

Muscle fatigue significantly impacts coordination, stability, and speed in daily activities. Accurate assessment of muscle fatigue is vital for effective exercise programs, injury prevention, and sports performance enhancement. Current methods mostly focus on individual muscles and strength evaluation, overlooking overall fatigue in multi-muscle movements. This study introduces a comprehensive muscle fatigue model using non-negative matrix factorization (NMF) weighting. NMF is employed to analyze the duration multi-muscle weight coefficient matrix (DMWCM) during synergistic movements, and four electromyographic (EMG) signal features in time, frequency, and complexity domains are selected. Particle Swarm Optimization (PSO) optimizes feature weights. The DMWCM and weighted features combine to calculate the Comprehensive Muscle Fatigue Index (CMFI) for multi-muscle synergistic movements. Experimental results show that CMFI correlates with perceived exertion (RPE) and Speed Dynamic Score (SDS), confirming its accuracy and real-time tracking in assessing multi-muscle synergistic movements. This model offers a more comprehensive approach to muscle fatigue assessment, with potential benefits for exercise training, injury prevention, and sports medicine.

Cholera is a leading cause of mortality in Nigeria. The two most significant predictors of cholera are a lack of access to clean water and poor sanitary conditions. Other factors such as natural disasters, illiteracy, and internal conflicts that drive people to seek sanctuary in refugee camps may contribute to the spread of cholera in Nigeria. The aim of this research is to develop a cholera outbreak risk prediction (CORP) model using machine learning tools and data science. In this study, we developed a CORP model using design science perspectives and machine learning to detect cholera outbreaks in Nigeria. Nonnegative matrix factorization (NMF) was used for dimensionality reduction, and synthetic minority oversampling technique (SMOTE) was used for data balancing. Outliers were detected using density-based spatial clustering of applications with noise (DBSCAN) were removed improving the overall performance of the model, and the extreme-gradient boost algorithm was used for prediction. The findings revealed that the CORP model outcomes resulted in the best accuracy of 99.62%, Matthews's correlation coefficient of 0.976, and area under the curve of 99.2%, which were improved compared with the previous findings. The developed model can be helpful to healthcare providers in predicting possible cholera outbreaks. HIGHLIGHTS Identifying the cholera prediction attributes.; Using socioeconomic variables to predict cholera.; The use of NMFSMOTE for outlier detection and balancing of the dataset.; Accuracy improved by the use of NMFSMOTE and DBSCAN.; The developed model can be helpful to healthcare providers in predicting possible cholera outbreaks.;

Pulsed eddy current thermography can detect surface or subsurface defects in steel, but in the process of combining deep learning, it is expensive and inefficient to build a complete sample of defects due to the complexity of the actual industrial environment. Consequently, this study proposes a transfer learning method based on Twin-NMF and combines it with the SimAM attention mechanism to enhance the detection accuracy of the target domain task. First, to address the domain differences between the target domain task and the source domain samples, this study introduces a Twin-NMF transfer method. This approach reconstructs the feature space of both the source and target domains using twin non-negative matrix factorization and employs cosine similarity to measure the correlation between the features of these two domains. Secondly, this study integrates a parameter-free SimAM into the neck of the YOLOv8 model to enhance its capabilities in extracting and classifying steel surface defects, as well as to alleviate the precision collapse phenomenon associated with multi-scale defect recognition. The experimental results show that the proposed Twin-NMF model with SimAM improves the detection accuracy of steel surface defects. Taking NEU-DET and GC10-DET as source domains, respectively, in the ECTI dataset, mAP@0.5 reaches 99.3% and 99.2%, and the detection accuracy reaches 98% and 98.5%.

The performance of any voice recognition platform in real environment depends on how well the desired speech signal is separated from unwanted signals like background noise or background speakers. In this paper, we propose a three stage hybrid model to separate two speakers from single channel speech mixture under unsupervised condition. Proposed method combines three techniques namely speech segmentation, NMF (Nonnegative Matrix Factorization) and Masking. Speech segmentation groups the short speech frames belonging to individual speakers by identifying the speaker change over points. The segmentation block groups the speech frames belonging to individual speakers but lacks in continuity of the speech samples. Therefore a second stage is built using NMF. NMF algorithm performs better in separating the speech mixture when parts of the individual speech signals are known a priori. This requirement is satisfied by speech segmentation stage. NMF further separates the individual speech signals in the mixture by maintaining continuity of speech samples over time. To further improve the accuracy of separated speech signals, various masking methods like TFR (Time frequency Ratio), SM (Soft Mask) and HM (Hard Mask) are applied. The separation results are compared with other unsupervised algorithms. The proposed hybrid model produces promising results in unsupervised single channel speech separation. This model can be applied at the front end of any voice recognition platform to further improve the recognition efficiency. [ABSTRACT FROM AUTHOR]

We propose an efficient distributed out-of-memory implementation of the non-negative matrix factorization (NMF) algorithm for heterogeneous high-performance-computing systems. The proposed implementation is based on prior work on NMFk, which can perform automatic model selection and extract latent variables and patterns from data. In this work, we extend NMFk by adding support for dense and sparse matrix operation on multi-node, multi-GPU systems. The resulting algorithm is optimized for out-of-memory problems where the memory required to factorize a given matrix is greater than the available GPU memory. Memory complexity is reduced by batching/tiling strategies, and sparse and dense matrix operations are significantly accelerated with GPU cores (or tensor cores when available). Input/output latency associated with batch copies between host and device is hidden using CUDA streams to overlap data transfers and compute asynchronously, and latency associated with collective communications (both intra-node and inter-node) is reduced using optimized NVIDIA Collective Communication Library (NCCL) based communicators. Benchmark results show significant improvement, from 32X to 76x speedup, with the new implementation using GPUs over the CPU-based NMFk. Good weak scaling was demonstrated on up to 4096 multi-GPU cluster nodes with approximately 25,000 GPUs when decomposing a dense 340 Terabyte-size matrix and an 11 Exabyte-size sparse matrix of density 10 - 6 . [ABSTRACT FROM AUTHOR]

Drug repurposing is a strategy for discovering new applications of existing drugs for use in various diseases. Despite the use of structured networks in drug research, it is still unclear how drugs interact with one another or with genes. Prostate adenocarcinoma is the second leading cause of cancer mortality in the United States, with an estimated incidence of 288,300 new cases and 34,700 deaths in 2023. In our study, we used integrative information from genes, pathways, and drugs for machine learning methods such as clustering, feature selection, and enrichment pathway analysis. We investigated how drugs affect drugs and how drugs affect genes in human pancreatic cancer cell lines that were derived from bone metastases of grade IV prostate cancer. Finally, we identified significant drug interactions within or between clusters, such as estradiol-rosiglitazone, estradiol-diclofenac, troglitazone-rosiglitazone, celecoxib-rofecoxib, celecoxib-diclofenac, and sodium phenylbutyrate-valproic acid. [ABSTRACT FROM AUTHOR]

Surface-enhanced Raman spectroscopy (SERS) has wide diagnostic applications due to narrow spectral features that allow multiplex analysis. We have previously developed a multiplexed, SERS-based nanosensor for micro-RNA (miRNA) detection called the inverse molecular sentinel (iMS). Machine learning (ML) algorithms have been increasingly adopted for spectral analysis due to their ability to discover underlying patterns and relationships within large and complex data sets. However, the high dimensionality of SERS data poses a challenge for traditional ML techniques, which can be prone to overfitting and poor generalization. Non-negative matrix factorization (NMF) reduces the dimensionality of SERS data while preserving information content. In this paper, we compared the performance of ML methods including convolutional neural network (CNN), support vector regression, and extreme gradient boosting combined with and without NMF for spectral unmixing of four-way multiplexed SERS spectra from iMS assays used for miRNA detection. CNN achieved high accuracy in spectral unmixing. Incorporating NMF before CNN drastically decreased memory and training demands without sacrificing model performance on SERS spectral unmixing. Additionally, models were interpreted using gradient class activation maps and partial dependency plots to understand predictions. These models were used to analyze clinical SERS data from single-plexed iMS in RNA extracted from 17 endoscopic tissue biopsies. CNN and CNN-NMF, trained on multiplexed data, performed most accurately with RMSElabel = 0.101 and 9.68 × 10–2, respectively. We demonstrated that CNN-based ML shows great promise in spectral unmixing of multiplexed SERS spectra, and the effect of dimensionality reduction on performance and training speed. Graphical abstract This is a visual representation of the abstract. [ABSTRACT FROM AUTHOR]

Cholera is a leading cause of mortality in Nigeria. The two most significant predictors of cholera are a lack of access to clean water and poor sanitary conditions. Other factors such as natural disasters, illiteracy, and internal conflicts that drive people to seek sanctuary in refugee camps may contribute to the spread of cholera in Nigeria. The aim of this research is to develop a cholera outbreak risk prediction (CORP) model using machine learning tools and data science. In this study, we developed a CORP model using design science perspectives and machine learning to detect cholera outbreaks in Nigeria. Nonnegative matrix factorization (NMF) was used for dimensionality reduction, and synthetic minority oversampling technique (SMOTE) was used for data balancing. Outliers were detected using density-based spatial clustering of applications with noise (DBSCAN) were removed improving the overall performance of the model, and the extreme-gradient boost algorithm was used for prediction. The findings revealed that the CORP model outcomes resulted in the best accuracy of 99.62%, Matthews's correlation coefficient of 0.976, and area under the curve of 99.2%, which were improved compared with the previous findings. The developed model can be helpful to healthcare providers in predicting possible cholera outbreaks. [ABSTRACT FROM AUTHOR]

As an efficient and intuitive dimension reduction algorithm, non-negative matrix factorization (NMF) has been widely employed in various fields. However, the existing NMF based methods have two disadvantages. Firstly, it treats each sample equally without considering the noise problem. Secondly, it does not restrict the coefficient matrix. Therefore, this paper proposes a novel weighted NMF algorithm based on adaptive robust local sparse graph (WNMF-ARLS), which includes the following superiorities compared with the other NMF-based algorithms: 1) The proposed method introduces a weighting regularization term, which distributes smaller weights to outliers and noise, and allocates larger weights to clean data. 2) Our method constructs a sparse local constraint graph to discover the data's potential manifold structure. 3) Unlike most NMF algorithms based on graph regularization, in which the graphs remain unchanged and are pre-defined during the NMF process, the proposed method introduces sparse constraints and local constraints into the unified framework to adaptively construct the optimization graph. Lots of image clustering experiments are provided to illustrate the effectiveness and superiority of the proposed WNMF-ARLS algorithm. Experimental results also show that the clustering performance of the proposed method is significantly better than that of other comparison algorithms. [ABSTRACT FROM AUTHOR]

Abstract Background With the emergence of metagenomic data, multiple links between the gut microbiome and the host health have been shown. Deciphering these complex interactions require evolved analysis methods focusing on the microbial ecosystem functions. Despite the fact that host or diet-derived fibres are the most abundant nutrients available in the gut, the presence of distinct functional traits regarding fibre and mucin hydrolysis, fermentation and hydrogenotrophic processes has never been investigated. Results After manually selecting 91 KEGG orthologies and 33 glycoside hydrolases further aggregated in 101 functional descriptors representative of fibre and mucin degradation pathways in the gut microbiome, we used nonnegative matrix factorization to mine metagenomic datasets. Four distinct metabolic profiles were further identified on a training set of 1153 samples, thoroughly validated on a large database of 2571 unseen samples from 5 external metagenomic cohorts and confirmed with metatranscriptomic data. Profiles 1 and 2 are the main contributors to the fibre-degradation-related metagenome: they present contrasted involvement in fibre degradation and sugar metabolism and are differentially linked to dysbiosis, metabolic disease and inflammation. Profile 1 takes over Profile 2 in healthy samples, and unbalance of these profiles characterize dysbiotic samples. Furthermore, high fibre diet favours a healthy balance between profiles 1 and profile 2. Profile 3 takes over profile 2 during Crohn’s disease, inducing functional reorientations towards unusual metabolism such as fucose and H2S degradation or propionate, acetone and butanediol production. Profile 4 gathers under-represented functions, like methanogenesis. Two taxonomic makes up of the profiles were investigated, using either the covariation of 203 prevalent genomes or metagenomic species, both providing consistent results in line with their functional characteristics. This taxonomic characterization showed that profiles 1 and 2 were respectively mainly composed of bacteria from the phyla Bacteroidetes and Firmicutes while profile 3 is representative of Proteobacteria and profile 4 of methanogens. Conclusions Integrating anaerobic microbiology knowledge with statistical learning can narrow down the metagenomic analysis to investigate functional profiles. Applying this approach to fibre degradation in the gut ended with 4 distinct functional profiles that can be easily monitored as markers of diet, dysbiosis, inflammation and disease. Video Abstract

Abstract Background Most partial hand amputees experience limited wrist movement. The limited rotational wrist movement deteriorates natural upper limb system related to hand use and the usability of the prosthetic hand, which may cause secondary damage to the musculoskeletal system due to overuse of the upper limb affected by repetitive compensatory movement patterns. Nevertheless, partial hand prosthetics, in common, have only been proposed without rotational wrist movement because patients have various hand shapes, and a prosthetic hand should be attached to a narrow space. Methods We hypothesized that partial hand amputees, when using a prosthetic hand with a wrist rotation module, would achieve natural upper limb movement muscle synergy and motion analysis comparable to a control group. To validate the proposed prototype design with the wrist rotation module and verify our hypothesis, we compared a control group with partial hand amputees wearing hand prostheses, both with and without the wrist rotation module prototype. The study contained muscle synergy analysis through non-negative matrix factorization (NMF) using surface electromyography (sEMG) and motion analyses employing a motion capture system during the reach-to-grasp task. Additionally, we assessed the usability of the prototype design for partial hand amputees using the Jebsen-Taylor hand function test (JHFT). Results The results showed that the number of muscle synergies identified through NMF remained consistent at 3 for both the control group and amputees using a hand prosthesis with a wrist rotation module. In the motion analysis, a statistically significant difference was observed between the control group and the prosthetic hand without the wrist rotation module, indicating the presence of compensatory movements when utilizing a prosthetic hand lacking this module. Furthermore, among the amputees, the JHFT demonstrated a greater improvement in total score when using the prosthetic hand equipped with a wrist rotation module compared to the prosthetic hand without this module. Conclusion In conclusion, integrating a wrist rotation module in prosthetic hand designs for partial hand amputees restores natural upper limb movement patterns, reduces compensatory movements, and prevent the secondary musculoskeletal. This highlights the importance of this module in enhancing overall functionality and quality of life.

This study introduces an unsupervised term extraction approach that combines non-negative matrix factorization (NMF) with word embeddings. Inspired by a pioneering semantic NMF method that employs regularization to jointly optimize document–word and word–word matrix factorizations for document clustering, we adapt this strategy for term extraction. Typically, a word–word matrix representing semantic relationships between words is constructed using cosine similarities between word embeddings. However, it has been established that transformer encoder embeddings tend to reside within a narrow cone, leading to consistently high cosine similarities between words. To address this issue, we replace the conventional word–word matrix with a word–seed submatrix, restricting columns to ‘domain seeds’—specific words that encapsulate the essential semantic features of the domain. Therefore, we propose a modified NMF framework that jointly factorizes the document–word and word–seed matrices, producing more precise encoding vectors for words, which we utilize to extract high-relevancy topic-related terms. Our modification significantly improves term extraction effectiveness, marking the first implementation of semantically enhanced NMF, designed specifically for the task of term extraction. Comparative experiments demonstrate that our method outperforms both traditional NMF and advanced transformer-based methods such as KeyBERT and BERTopic. To support further research and application, we compile and manually annotate two new datasets, each containing 1000 sentences, from the ‘Geography and History’ and ‘National Heroes’ domains. These datasets are useful for both term extraction and document classification tasks. All related code and datasets are freely available.

The choice of the factorization rank of a matrix is critical, e.g., in dimensionality reduction, filtering, clustering, deconvolution, etc., because selecting a rank that is too high amounts to adjusting the noise, while selecting a rank that is too low results in the oversimplification of the signal. Numerous methods for selecting the factorization rank of a non-negative matrix have been proposed. One of them is the cophenetic correlation coefficient ( c c c ), widely used in data science to evaluate the number of clusters in a hierarchical clustering. In previous work, it was shown that c c c performs better than other methods for rank selection in non-negative matrix factorization (NMF) when the underlying structure of the matrix consists of orthogonal clusters. In this article, we show that using the ratio of c c c to the approximation error significantly improves the accuracy of the rank selection. We also propose a new criterion, c o n c o r d a n c e , which, like c c c , benefits from the stochastic nature of NMF; its accuracy is also improved by using its ratio-to-error form. Using real and simulated data, we show that c o n c o r d a n c e , with a CUSUM-based automatic detection algorithm for its original or ratio-to-error forms, significantly outperforms c c c . It is important to note that the new criterion works for a broader class of matrices, where the underlying clusters are not assumed to be orthogonal. [ABSTRACT FROM AUTHOR]

In this paper, we propose a genetic algorithm-based adaptive Wiener gain for speech enhancement using an iterative posterior non-negative matrix factorization (NMF). In the recent past, NMF-based Wiener filtering methods were used to improve the performance of speech enhancement, which has shown that they provide better performance when compared with conventional NMF methods. But performance degrades in non-stationary noise environments. Template-based approaches are more robust and perform better in non-stationary noise environments compared to statistical model-based approaches but are dependent on a priori information. Combining the approaches avoids the drawbacks of both. To improve the performance further, speech and noise bases are adapted simultaneously in the NMF approach. The usage of Super-Gaussian constraints in iterative NMF still improves the performance in non-stationary noise. The silence frame is a challenging task in the case of NMF; still there will be some amount of noise present in those frames. For further enhancement, we have combined with a genetic algorithm (GA)-based adaptive Wiener filter which performs well in denoising and also the GA search the adaptive α allows us to control the trade-off between fitting the observed spectrogram of mixed speech and noise achieving high likelihood under our prior model. The proposed method outperforms other benchmark algorithms in terms of the source to distortion ratio (SDR), short-time objective intelligibility (STOI), and perceptual evaluation of speech quality (PESQ). [ABSTRACT FROM AUTHOR]