Application Of Non-negative Matrix Factorization Research Articles

Predicting the potential associations between circRNA and drug sensitivity using a multisource feature-based approach.

The unique non-coding RNA molecule known as circular RNA (circRNA) is distinguished from conventional linear RNA by having a longer half-life, greater degree of conservation and inherent solidity. Extensive research has demonstrated the profound impact of circRNA expression on cellular drug sensitivity and therapeutic efficacy. There is an immediate need for the creation of efficient computational techniques to anticipate the potential correlations between circRNA and drug sensitivity, as classical biological research approaches are time-consuming and costly. In this work, we introduce a novel deep learning model called SNMGCDA, which aims to forecast the relationships between circRNA and drug sensitivity. SNMGCDA incorporates a diverse range of similarity networks, enabling the derivation of feature vectors for circRNAs and drugs using three distinct calculation methods. First, we utilize a sparse autoencoder for the extraction of drug characteristics. Subsequently, the application of non-negative matrix factorization (NMF) enables the identification of relationships between circRNAs and drugs based on their shared features. Additionally, the multi-head graph attention network is employed to capture the characteristics of circRNAs. After acquiring the characteristics from these three separate components, we combine them to form a unified and inclusive feature vector for each cluster of circRNA and drug. Finally, the relevant feature vectors and labels are inputted into a multilayer perceptron (MLP) to make predictions. The outcomes of the experiment, obtained through 5-fold cross-validation (5-fold CV) and 10-fold cross-validation (10-fold CV), demonstrate SNMGCDA outperforms five other state-of-art methods in terms of performance. Additionally, the majority of case studies have predominantly confirmed newly discovered correlations by SNMGCDA, thereby emphasizing its reliability in predicting potential relationships between circRNAs and drugs.

Progression from cardiomyopathy to heart failure with reduced ejection fraction: A CORIN deficient course

Cardiomyopathies, encompassing hypertrophic cardiomyopathy (HCM) and dilated cardiomyopathy (DCM), constitute a diverse spectrum of heart muscle diseases that often culminating in heart failure (HF). The inherent molecular heterogeneity of these conditions has implications for prognosis and therapeutic strategies.Publicly available microarray and RNA sequencing (RNA-seq) data sets of HCM (n = 106 from GSE36961) and DCM (n = 18 from GSE135055 and 166 from GSE141910) patients were employed for our analysis. The Non-negative Matrix Factorization (NMF) algorithm was applied to explore the molecular stratification within HCM and DCM, and enrichment analysis was performed to delineate their biological characteristics. By integrating bulk and single-nucleus RNA-seq (snRNA-seq) data, we identified a potential biomarker for HCM progression and cardiac fibrosis, which was subsequently validated using mendelian randomization and in vitro.Our application of NMF identified two distinct molecular clusters. Particularly, a profibrotic, heart failure with reduced ejection fraction (HFrEF)-resembling Cluster 1 emerged, characterized by diminished expression of CORIN and a high degree of fibroblast activation. This cluster also exhibited lower left ventricular ejection fraction (LVEF) and worse prognostic outcomes, establishing the significance of this molecular subclassification. We further found that overexpression of CORIN could mitigate TGFβ1-induced expression of col1a1 and α-SMA in neonatal rat cardiac fibroblasts.Our results indicated the heterogeneity of HCM population, and further evidenced the participation of corin in the progression of HCM, DCM and HFrEF. Nevertheless, our study is constrained by the lack of corresponding clinical data and experimental validation of the identified subtypes. Therefore, further studies are warranted to elucidate the downstream pathways of corin and to validate these findings in independent patient cohorts.

Application of Non-negative Matrix Factorization in Processing Self-potential Signals Caused by Geo-microbe

Microorganisms play a significant role in the evolutionary process of the Earth and are closely related to the modification and evolution of the Earth’s surface circles at different spatial and temporal scales. Given geological microbiology in the self-potential signal data can be represented with non-negative matrix, which makes use of non-negative matrix factorization method to deal with the data become possible. This paper introduces the basic principle of non-negative matrix factorization method, based on non-negative matrix factorization method is discussed to isolate the self-potential signal data in geological microbes may and effect. By separating two natural potential signals containing noise with blind sources, the separation effect is good and the non-negative matrix factorization method is found to be an effective method for extracting their self-potential signal data. Although this numerical simulation is the self-potential signal is introduced in the work of geological microbiology, but this method also has reference significance for other signal data processing, can also be used to other geophysical data processing, the implementation of the research on environmental protection, to improve the cognitive complex biological geological process of shallow earth environment has important significance.

Identification an innovative classification and nomogram for predicting the prognosis of thyroid carcinoma patients and providing therapeutic schedules.

Thyroid carcinoma (THCA) represents a prevalent form of cancer globally, with its incidence demonstrating an upward trend in recent years. Accumulating evidence has indicated that programmed cell death (PCD) patterns exert a vital influence on tumor progression. Nevertheless, the association between PCD and the prognosis of patients with papillary thyroid carcinoma remains to be elucidated. The current study endeavors to examine the link between PCD and the prognosis of thyroid cancer while concurrently developing a prognostic index based on PCD genes. Programmed cell death patterns were employed to construct the model and define clusters. Gene expression profile genomics and clinical data pertaining to 568 patients with thyroid cancer were sourced from the TCGA database. In addition, single-cell transcriptome data GSE184362 were procured from the Gene Expression Omnibus (GEO) database for subsequent analysis. The study harnessed six machine learning algorithms to create a programmed cell death signature (PCDS). Ultimately, the model developed via SVM was chosen as the optimal model, boasting the highest C-index. Moreover, the application of non-negative matrix factorization (NMF) led to the identification of two molecular subtypes of THCA, each characterized by distinct vital biological processes and drug sensitivities. The investigation revealed that PCDS is linked to chemokines, interleukins, interferons, and checkpoint genes, as well as pivotal components of the tumor microenvironment, as determined through a comprehensive analysis of bulk and single-cell transcriptomes. Patients with THCA and elevated PCDS values are more inclined to exhibit resistance to conventional chemotherapy regimens, yet may display heightened responsiveness to targeted therapeutic agents. Finally, we established a nomogram model based on multivariable cox and logistic regression analyses to predict the overall survival of THCA patients. This research sheds new light on the role of programmed cell death (PCD) patterns in THCA. By conducting an in-depth analysis of various cell death patterns, a novel PCD model has been devised, capable of accurately predicting the clinical prognosis and drug sensitivity of patients with THCA.

A new gene-scoring method for uncovering novel glaucoma-related genes using non-negative matrix factorization based on RNA-seq data.

Early diagnosis and treatment of glaucoma are challenging. The discovery of glaucoma biomarkers based on gene expression data could potentially provide new insights for early diagnosis, monitoring, and treatment options of glaucoma. Non-negative Matrix Factorization (NMF) has been widely used in numerous transcriptome data analyses in order to identify subtypes and biomarkers of different diseases; however, its application in glaucoma biomarker discovery has not been previously reported. Our study applied NMF to extract latent representations of RNA-seq data from BXD mouse strains and sorted the genes based on a novel gene scoring method. The enrichment ratio of the glaucoma-reference genes, extracted from multiple relevant resources, was compared using both the classical differentially expressed gene (DEG) analysis and NMF methods. The complete pipeline was validated using an independent RNA-seq dataset. Findings showed our NMF method significantly improved the enrichment detection of glaucoma genes. The application of NMF with the scoring method showed great promise in the identification of marker genes for glaucoma.

NMF Clustering: Accessible NMF-based Clustering Utilizing GPU Acceleration.

Non-negative Matrix Factorization (NMF) is an algorithm that can reduce high dimensional datasets of tens of thousands of genes to a handful of metagenes which are biologically easier to interpret. Application of NMF on gene expression data has been limited by its computationally intensive nature, which hinders its use on large datasets such as single-cell RNA sequencing (scRNA-seq) count matrices. We have implemented NMF based clustering to run on high performance GPU compute nodes using CuPy, a GPU backed python library, and the Message Passing Interface (MPI). This reduces the computation time by up to three orders of magnitude and makes the NMF Clustering analysis of large RNA-Seq and scRNA-seq datasets practical. We have made the method freely available through the GenePattern gateway, which provides free public access to hundreds of tools for the analysis and visualization of multiple 'omic data types. Its web-based interface gives easy access to these tools and allows the creation of multi-step analysis pipelines on high performance computing (HPC) clusters that enable reproducible in silico research for non-programmers.

Application of non-negative matrix factorization in oncology: one approach for establishing precision medicine.

The increase in the expectations of artificial intelligence (AI) technology has led to machine learning technology being actively used in the medical field. Non-negative matrix factorization (NMF) is a machine learning technique used for image analysis, speech recognition, and language processing; recently, it is being applied to medical research. Precision medicine, wherein important information is extracted from large-scale medical data to provide optimal medical care for every individual, is considered important in medical policies globally, and the application of machine learning techniques to this end is being handled in several ways. NMF is also introduced differently because of the characteristics of its algorithms. In this review, the importance of NMF in the field of medicine, with a focus on the field of oncology, is described by explaining the mathematical science of NMF and the characteristics of the algorithm, providing examples of how NMF can be used to establish precision medicine, and presenting the challenges of NMF. Finally, the direction regarding the effective use of NMF in the field of oncology is also discussed.

Prognostic mutational subtyping in de novo diffuse large B-cell lymphoma

BackgroundDiffuse large B-cell lymphoma (DLBCL) is a heterogeneous disease defined using a number of well-established molecular subsets. Application of non-negative matrix factorization (NMF) to whole exome sequence data has previously been used to identify six distinct molecular clusters in DLBCL with potential clinical relevance. In this study, we applied NMF-clustering to targeted sequencing data utilizing the FoundationOne Heme® panel from the Phase III GOYA (NCT01287741) and Phase Ib/II CAVALLI studies (NCT02055820) in de novo DLBCL. Biopsy samples, survival outcomes, RNA-Seq and targeted exome-sequencing data were available for 423 patients in GOYA (obinutuzumab [G]-cyclophosphamide, doxorubicin, vincristine, and prednisone [CHOP] vs rituximab [R]-CHOP) and 86 patients in CAVALLI (venetoclax+[G/R]-CHOP).ResultsWhen the NMF algorithm was applied to samples from the GOYA study analyzed using a comprehensive genomic profiling platform, four of the six groups previously reported were observed: MYD88/CD79B, BCL2/EZH2, NOTCH2/TNFAIP3, and no mutations. Mutation profiles, cell-of-origin subset distributions and clinical associations of MYD88/CD79B and BCL2/EZH2 groups were similar to those described in previous NMF studies. In contrast, application of NMF to the CAVALLI study yielded only three; MYD88/CD79B-, BCL2/EZH2-like clusters, and a no mutations group, and there was a trend towards improved outcomes for BCL2/EZH2 over MYD88/CD79B.ConclusionsThis analysis supports the utility of NMF used in conjunction with targeted sequencing platforms for identifying patients with different prognostic subsets. The observed trend for improved overall survival in the BCL2/EZH2 group is consistent with the mechanism of action of venetoclax, suggesting that targeting sequencing and NMF has potential for identifying patients who are more likely to gain benefit from venetoclax therapy.

Detecting cell assemblies by NMF-based clustering from calcium imaging data

A large number of neurons form cell assemblies that process information in the brain. Recent developments in measurement technology, one of which is calcium imaging, have made it possible to study cell assemblies. In this study, we aim to extract cell assemblies from calcium imaging data. We propose a clustering approach based on non-negative matrix factorization (NMF). The proposed approach first obtains a similarity matrix between neurons by NMF and then performs spectral clustering on it. The application of NMF entails the problem of model selection. The number of bases in NMF affects the result considerably, and a suitable selection method is yet to be established. We attempt to resolve this problem by model averaging with a newly defined estimator based on NMF. Experiments on simulated data suggest that the proposed approach is superior to conventional correlation-based clustering methods over a wide range of sampling rates. We also analyzed calcium imaging data of sleeping/waking mice and the results suggest that the size of the cell assembly depends on the degree and spatial extent of slow wave generation in the cerebral cortex.

GEOMETRIC STRUCTURE GUIDED MODEL AND ALGORITHMS FOR COMPLETE DECONVOLUTION OF GENE EXPRESSION DATA.

Complete deconvolution analysis for bulk RNA-seq data is important and helpful to distinguish whether the differences of disease-associated GEPs (gene expression profiles) in tissues of patients and normal controls are due to changes in cellular composition of tissue samples, or due to GEPs changes in specific cells. One of the major techniques to perform complete deconvolution is nonnegative matrix factorization (NMF), which also has a wide-range of applications in the machine learning community. However, the NMF is a well-known strongly ill-posed problem, so a direct application of NMF to RNA-seq data will suffer severe difficulties in the interpretability of solutions. In this paper, we develop an NMF-based mathematical model and corresponding computational algorithms to improve the solution identifiability of deconvoluting bulk RNA-seq data. In our approach, we combine the biological concept of marker genes with the solvability conditions of the NMF theories, and develop a geometric structures guided optimization model. In this strategy, the geometric structure of bulk tissue data is first explored by the spectral clustering technique. Then, the identified information of marker genes is integrated as solvability constraints, while the overall correlation graph is used as manifold regularization. Both synthetic and biological data are used to validate the proposed model and algorithms, from which solution interpretability and accuracy are significantly improved.

A Novel Application of Non-Negative Matrix Factorization to the Prediction of the Health Status of Undocumented Immigrants.

Introduction: Undocumented immigrants (UIs) in the United States are less likely to be able to afford health insurance. As a result, UIs often lack family doctors and are rarely involved in annual screening programs, which makes estimating their health status remarkably challenging. This is especially true if the laboratory results from limited screening programs fail to provide sufficient clinical information.Methods: To address this issue, we have developed a machine learning model based on the non-negative matrix factorization technique. The data set we used for model training and testing was obtained from the 2004 cost-free hepatitis B screening program at the Omni Health Center located in Plano, Texas. Total 300 people were involved, with 199 identified as UIs.Results: People in the UIs group have higher cholesterol (219.6 mg/dL, p=0.038) and triglycerides (173.2 mg/dL, p=0.03) level. They also have a lower hepatitis B vaccination rate (38%, p=0.0247). No significant difference in hepatitis B(+) was found (p=0.8823). Using 16 individual clinical measurements as training features, our newly developed model has a 67.56% accuracy in predicting the ratio of cholesterol to high-density lipoprotein; in addition, this newly developed model performs 9.1% better than a comparable multiclass logistic regression model.Conclusions: Elderly UIs have poorer health status compared with permanent residents and citizens in the United States. Our newly developed machine learning model demonstrates a powerful support tool for designing health intervention programs that target UIs in the United States.

Automatic estimation of the sound emergence of wind turbine noise with nonnegative matrix factorization.

In many countries, the acoustic impact of wind farms is often constrained by a curtailment plan to limit their noise, which spreads in their surroundings. To update the plan, on/off cycle measurements are performed to determine the ambient noise (wind turbines in operation) and residual noise (wind turbines shut down), but these shutdown operations are limited in time, which reduces the representativeness of the estimated in situ emergence. Consequently, a machine learning technique, called nonnegative matrix factorization (NMF), is proposed to estimate the sound emergence of wind turbines continuously, i.e., without stopping the machines. In the first step, the application of NMF on a corpus of various simulated scenes allows the determination of the optimal setting of the method to better estimate the sound emergence. The results show the proper adaptation of the method with regard to the influence of the propagation distance and atmospheric conditions. This method also proves to be efficient in cases in which the real emergence is less than 5 dB(A) with a mean error lower than 2 dB(A). The first comparison with in situ measurements validates these performances and allows the consideration of the application of this method to optimize wind farm operations.

Tracking Proterozoic–Triassic sediment routing to western Laurentia via bivariate non-negative matrix factorization of detrital provenance data

When reconstructing sediment provenance, the challenge posed by multiple univariate (i.e., age-only) detrital geochronology data sets with similar one-dimensional distributions can be mitigated by incorporation of a second variable such as Hf isotopic data. However, there is no commonly applied method of sample intercomparison, much less forward or inverse modeling of these bivariate data sets. In this paper we explore application of non-negative matrix factorization (NMF) to bivariate data sets. Factorization of univariate mixed (a.k.a., sink or daughter) data sets has been demonstrated to successfully recover both the one-dimensional endmember (a.k.a., source or parent) distributions and their mixture weightings. We show that NMF can successfully recover both the two-dimensional distributions and mixing weights in synthetic data sets. Application of the method to 24 published Neoproterozoic–Triassic samples from western Laurentia yields six two-dimensional endmember distributions that are a close match to empirical sediment sources on the southern, eastern, and northern margin of Laurentia. The results are broadly consistent with previous interpretations and confirm that the method can characterize unknown sediment sources based on data from analyzed sink samples. Correlating between factorized endmembers and empirical sources indicates that the Transcontinental Arch was not a barrier to east–west sediment transport until the late Cambrian. Quantitative comparison also shows that the closest known match for a factorized endmember with highly evolved Permo-Triassic zircons is from northern South America and suggests northward transport of this detritus following assembly of Pangea. Recognizing these empirical sources and their distribution into strata that would later be incorporated into the North American Cordillera sets the stage for interpreting sediment provenance records in the Jurassic–Paleogene Cordilleran retroarc foreland basin. Supplementary material: The open-source software ( DZnmf2D ) and a user manual is available in public repositories: https://doi.org/10.5281/zenodo.4460336 . All other supplemental files are available at https://doi.org/10.6084/m9.figshare.c.5280320 Thematic collection: This article is part of the Fold-and-thrust belts and associated basins collection available at: https://www.lyellcollection.org/cc/fold-and-thrust-belts

Monitor Ionizing Radiation-Induced Cellular Responses with Raman Spectroscopy, Non-Negative Matrix Factorization, and Non-Negative Least Squares

Radiation therapy (RT) is one of the most commonly prescribed cancer treatments. New tools that can accurately monitor and evaluate individual patient responses would be a major advantage and lend to the implementation of personalized treatment plans. In this study, Raman spectroscopy (RS) was applied to examine radiation-induced cellular responses in H460, MCF7, and LNCaP cancer cell lines across different dose levels and times post-irradiation. Previous Raman data analysis was conducted using principal component analysis (PCA), which showed the ability to extract biological information of glycogen. In the current studies, the use of non-negative matrix factorization (NMF) allowed for the discovery of multiplexed biological information, specifically uncovering glycogen-like and lipid-like component bases. The corresponding scores of glycogen and previously unidentified lipids revealed the content variations of these two chemicals in the cellular data. The NMF decomposed glycogen and lipid-like bases were able to separate the cancer cell lines into radiosensitive and radioresistant groups. A further lipid phenotype investigation was also attempted by applying non-negative least squares (NNLS) to the lipid-like bases decomposed individually from three cell lines. Qualitative differences found in lipid weights for each lipid-like basis suggest the lipid phenotype differences in the three tested cancer cell lines. Collectively, this study demonstrates that the application of NMF and NNLS on RS data analysis to monitor ionizing radiation-induced cellular responses can yield multiplexed biological information on bio-response to RT not revealed by conventional chemometric approaches.

Active Set Type Algorithms for Nonnegative Matrix Factorization in Hyperspectral Unmixing

Hyperspectral unmixing is a powerful method of the remote sensing image mining that identifies the constituent materials and estimates the corresponding fractions from the mixture. We consider the application of nonnegative matrix factorization (NMF) for the mining and analysis of spectral data. In this paper, we develop two effective active set type NMF algorithms for hyperspectral unmixing. Because the factor matrices used in unmixing have sparse features, the active set strategy helps reduce the computational cost. These active set type algorithms for NMF is based on an alternating nonnegative constrained least squares (ANLS) and achieve a quadratic convergence rate under the reasonable assumptions. Finally, numerical tests demonstrate that these algorithms work well and that the function values decrease faster than those obtained with other algorithms.

Recapitulation of Prognostic Mutational Subtyping Utilizing F1H in GOYA and CAVALLI and the Potential to Highlight Benefit of Targeted Therapy in De Novo DLBCL

Introduction: Diffuse large B-cell lymphoma (DLBCL), is a genetically and clinically heterogeneous disease. Building on existing clinical prognostic scores in the era of genomic risk modeling is critical to improve patient (pt) outcomes and inform future trials. With advances in next-generation sequencing, the interrogation of genetic drivers to refine prognostic subgroups over and above histological cell of origin (COO) subtypes has been of keen interest. The application of an unsupervised prognostic clustering approach, non-negative matrix factorization (NMF) algorithm, to whole exome sequence (WES) data has led to identification of six distinct molecular clusters with potential clinical relevance (Chapuy et al. Nature Med, 2018). Here, we applied the NMF clustering approach to targeted exome-sequencing data utilizing the FoundationOne HemeTM panel and analyzed RNA gene expression data from the randomized Phase 3 GOYA and Phase 1b/2 CAVALLI studies in de novo DLBCL. Methods: Biopsy samples, clinical outcomes (overall survival [OS], progression-free survival [PFS]), and targeted exome-sequencing data were available for 499 pts from the GOYA trial (NCT01287741; ITT=1418; obinutuzumab [G]-CHOP vs rituximab [R]-CHOP) and 106 pts from the CAVALLI trial (NCT02055820; ITT=267, venetoclax+[G/R]-CHOP). The FoundationOne HemeTM panel of 465 genes (F1H, Foundation Medicine Inc. [FMI]) was used to define single nucleotide variants, copy number amplifications, and rearrangements. Clustering methodology was implemented using the R package NMF (Gaujoux and Seoighe, BMC Bioinformatics 2010). All models were adjusted for treatment (where applicable), age, International Prognostic Index (IPI), sex, COO (NanoString), and BCL2 by IHC. Results: Applying the NMF algorithm, de novo pts in GOYA clustered similarly into four of the six groups recently reported by Chapuy et al.: MYD88/CD79B, BCL2/EZH2, NOTCH2/TNFAIP3, and no mutations (Fig 1A). The mutation profiles and COO subset distribution were also similar to those previously described. Furthermore, the MYD88/CD79B and BCL2/EZH2 groups remained associated with inferior clinical outcomes (Fig 1B), with observed poor prognosis over the NOTCH2/TNFAIP3 and no mutations groups (OS: hazard ratio [HR], 2.0; 95% CI: 0.97, 3.90; PFS: HR, 1.60; 95% CI: 0.93, 2.80), even when adjusting for known prognostic factors. Application of NMF to pts treated with the BCL2-inhibitor venetoclax in combination with R-CHOP (CAVALLI) yielded only MYD88/CD79B- and BCL2/EZH2-like groups due to enrollment. However, in contrast to the BCL2/EZH2 subgroup reported in both Chapuy et al. and GOYA, these pts demonstrated a trend towards improved outcomes over MYD88/CD79B (OS: HR, 0.12; 95% CI: 0.01, 1.60; PFS: HR, 0.11; 95% CI: 0.01, 1.10; Fig 1D). The difference between studies is potentially explained by the ability of venetoclax treatment to overcome the adverse prognostic impact of BCL2 over-expression. BCL2 IHC positive (BCL2+) pts were enriched in high-risk groups in GOYA (MYD88/CD79B and BCL2/EZH2) (odds ratio [OR], 5.0; 95% CI: 3.13, 8.33) and CAVALLI (MYD88/CD79B) (OR, 2.0; 95% CI: 0.63, 6.67). In GOYA, a prognostic trend for BCL2+ status is observed within low-risk pts (OS BCL2+ HR, 1.9; 95% CI: 0.55, 6.3; PFS BCL2+ HR, 2.7; 95% CI: 1.0, 6.8), but not within high-risk groups (OS BCL2+ HR, 0.55; 95% CI: 0.29, 1.00; PFS BCL2+ HR, 0.76; 95% CI: 0.45, 1.30; Fig 1C). Additional genetic risk models will be presented, e.g. as per Reddy et al. (Cell, 2017), as well as differential expression analysis. Conclusions: Important findings from the present work include: 1) utility of targeted mutational data from F1H in recapitulating molecular clusters, previously identified using WES, that further subclassify established prognostic groups such as COO and IPI; and 2) a trend towards improved survival outcomes for patients in the BCL2/EZH2 cluster upon treatment with the BCL2 inhibitor venetoclax. This highlights the potential for utilizing novel genetic signatures as a means of identifying patients suitable for targeted therapies in the era of personalized health care. Disclosures Kim: Genentech, Inc.: Employment. Fan:Genentech, Inc.: Employment. Bolen:F. Hoffmann-La Roche: Equity Ownership; Genentech, Inc.: Employment. Bazeos:Roche Products Ltd: Employment, Equity Ownership. Jiang:Genentech: Employment, Equity Ownership; F. Hoffman-La Roche: Equity Ownership. Balasubramanian:Roche-Genentech: Employment; F. Hoffman-La Roche: Equity Ownership. Balakrishnan:Genentech, Inc.: Employment. Knapp:F. Hoffmann-La Roche Ltd: Employment. Humphrey:F. Hoffmann-La Roche Ltd: Employment. Nielsen:F. Hoffmann-La Roche Ltd: Employment, Equity Ownership. Venstrom:F. Hoffmann-La Roche Ltd: Employment. Hernandez:Genentech, Inc.: Employment, Equity Ownership. Paulson:F. Hoffman-La Roche: Equity Ownership; Genentech: Employment.

Investigating microstructural variation in the human hippocampus using non-negative matrix factorization

In this work we use non-negative matrix factorization to identify patterns of microstructural variance in the human hippocampus. We utilize high-resolution structural and diffusion magnetic resonance imaging data from the Human Connectome Project to query hippocampus microstructure on a multivariate, voxelwise basis. Application of non-negative matrix factorization identifies spatial components (clusters of voxels sharing similar covariance patterns), as well as subject weightings (individual variance across hippocampus microstructure). By assessing the stability of spatial components as well as the accuracy of factorization, we identified 4 distinct microstructural components. Furthermore, we quantified the benefit of using multiple microstructural metrics by demonstrating that using three microstructural metrics (T1-weighted/T2-weighted signal, mean diffusivity and fractional anisotropy) produced more stable spatial components than when assessing metrics individually. Finally, we related individual subject weightings to demographic and behavioural measures using a partial least squares analysis. Through this approach we identified interpretable relationships between hippocampus microstructure and demographic and behavioural measures. Taken together, our work suggests non-negative matrix factorization as a spatially specific analytical approach for neuroimaging studies and advocates for the use of multiple metrics for data-driven component analyses.

Source separation and apportionment of surface water pollution in the Luanhe River Basin based on non-negative matrix factorization

Abstract Understanding the spatial and temporal variations and source apportionment of water pollution is important for efficient water environment management. The non-negative matrix factorization (NMF) method, which is naturally well suited for non-negative data of high dimension, was used to identify the latent factors and apportion the contributions from identified pollution sources to each water quality parameter. We obtained a data matrix with 11 water quality variables collected from 2013 to 2016 in the Luanhe River Basin in northern China. The results highlight the substantial contribution of industrial and livestock wastewater. All land-use types have a slightly weaker impact on surface water pollution during the dry season than during the rainy season. The aim of this study is to illustrate the practicability of multivariate statistical analysis, especially the application of NMF, which has major potential for source separation and the apportionment of water pollution.

Performance Evaluation of Selected Cost Functions in Non Negative Matrix Factorization Based Decomposition of Acoustic Mixture

Interaction of acoustic signals when several audio sources are active simultaneously results in the disturbance of estimation of an individual source by co-occurring sounds. Data decomposition therefore constitutes one of the core tasks in monaural source separation. Particularly, in semi-supervised learning approach, viable means of achieving this is through the application of Non-negative Matrix Factorization (NMF). Owing to a paucity of information on the application of this method, especially in a speech system, evaluation of some cost functions in NMF-based monaural speech decomposition was investigated in this study. A generalized gradient descent algorithm is derived for the minimization while three cost functions: Euclidean Distance, Kullback-Leibler Divergence and Itakura-Saito divergences are applied to the derived separation NMF algorithm. These divergences are evaluated using experimental data while the performance of each of these is evaluated based on the cost values and convergence rate. Itakura-Saito divergence yields optimal performance over the other two divergences for given number of iterations and number of channels. Keywords— Cost functions, non-negative matrix factorization, speech separation, evaluation

非负矩阵分解在空间目标图像识别中的应用

非负矩阵分解在空间目标图像识别中的应用