Visualization Of High-dimensional Data Research Articles

Visualizing High-Dimensional Data: Advances in the Past Decade

Massive simulations and arrays of sensing devices, in combination with increasing computing resources, have generated large, complex, high-dimensional datasets used to study phenomena across numerous fields of study. Visualization plays an important role in exploring such datasets. We provide a comprehensive survey of advances in high-dimensional data visualization that focuses on the past decade. We aim at providing guidance for data practitioners to navigate through a modular view of the recent advances, inspiring the creation of new visualizations along the enriched visualization pipeline, and identifying future opportunities for visualization research.

Mapping the Fetomaternal Peripheral Immune System at Term Pregnancy.

Preterm labor and infections are the leading causes of neonatal deaths worldwide. During pregnancy, immunological cross talk between the mother and her fetus is critical for the maintenance of pregnancy and the delivery of an immunocompetent neonate. A precise understanding of healthy fetomaternal immunity is the important first step to identifying dysregulated immune mechanisms driving adverse maternal or neonatal outcomes. This study combined single-cell mass cytometry of paired peripheral and umbilical cord blood samples from mothers and their neonates with a graphical approach developed for the visualization of high-dimensional data to provide a high-resolution reference map of the cellular composition and functional organization of the healthy fetal and maternal immune systems at birth. The approach enabled mapping of known phenotypical and functional characteristics of fetal immunity (including the functional hyperresponsiveness of CD4+ and CD8+ T cells and the global blunting of innate immune responses). It also allowed discovery of new properties that distinguish the fetal and maternal immune systems. For example, examination of paired samples revealed differences in endogenous signaling tone that are unique to a mother and her offspring, including increased ERK1/2, MAPK-activated protein kinase 2, rpS6, and CREB phosphorylation in fetal Tbet+CD4+ T cells, CD8+ T cells, B cells, and CD56loCD16+ NK cells and decreased ERK1/2, MAPK-activated protein kinase 2, and STAT1 phosphorylation in fetal intermediate and nonclassical monocytes. This highly interactive functional map of healthy fetomaternal immunity builds the core reference for a growing data repository that will allow inferring deviations from normal associated with adverse maternal and neonatal outcomes.

SCALABLE VISUALIZATION FOR HIGH-DIMENSIONAL SINGLE-CELL DATA.

Single-cell analysis can uncover the mysteries in the state of individual cells and enable us to construct new models about the analysis of heterogeneous tissues. State-of-the-art technologies for single-cell analysis have been developed to measure the properties of single-cells and detect hidden information. They are able to provide the measurements of dozens of features simultaneously in each cell. However, due to the high-dimensionality, heterogeneous complexity and sheer enormity of single-cell data, its interpretation is challenging. Thus, new methods to overcome high-dimensionality are necessary. Here, we present a computational tool that allows efficient visualization of high-dimensional single-cell data onto a low-dimensional (2D or 3D) space while preserving the similarity structure between single-cells. We first construct a network that can represent the similarity structure between the high-dimensional representations of single-cells, and then, embed this network into a low-dimensional space through an efficient online optimization method based on the idea of negative sampling. Using this approach, we can preserve the high-dimensional structure of single-cell data in an embedded low-dimensional space that facilitates visual analyses of the data.

고차원 대용량 자료의 시각화에 대한 고찰

고차원 대용량 자료의 시각화에 대한 고찰

Approximated and User Steerable tSNE for Progressive Visual Analytics.

Progressive Visual Analytics aims at improving the interactivity in existing analytics techniques by means of visualization as well as interaction with intermediate results. One key method for data analysis is dimensionality reduction, for example, to produce 2D embeddings that can be visualized and analyzed efficiently. t-Distributed Stochastic Neighbor Embedding (tSNE) is a well-suited technique for the visualization of high-dimensional data. tSNE can create meaningful intermediate results but suffers from a slow initialization that constrains its application in Progressive Visual Analytics. We introduce a controllable tSNE approximation (A-tSNE), which trades off speed and accuracy, to enable interactive data exploration. We offer real-time visualization techniques, including a density-based solution and a Magic Lens to inspect the degree of approximation. With this feedback, the user can decide on local refinements and steer the approximation level during the analysis. We demonstrate our technique with several datasets, in a real-world research scenario and for the real-time analysis of high-dimensional streams to illustrate its effectiveness for interactive data analysis.

Semantic Visualization with Neighborhood Graph Regularization

Visualization of high-dimensional data, such as text documents, is useful to map out the similarities among various data points. In the high-dimensional space, documents are commonly represented as bags of words, with dimensionality equal to the vocabulary size. Classical approaches to document visualization directly reduce this into visualizable two or three dimensions. Recent approaches consider an intermediate representation in topic space, between word space and visualization space, which preserves the semantics by topic modeling. While aiming for a good fit between the model parameters and the observed data, previous approaches have not considered the local consistency among data instances. We consider the problem of semantic visualization by jointly modeling topics and visualization on the intrinsic document manifold, modeled using a neighborhood graph. Each document has both a topic distribution and visualization coordinate. Specifically, we propose an unsupervised probabilistic model, called SEMAFORE, which aims to preserve the manifold in the lower-dimensional spaces through a neighborhood regularization framework designed for the semantic visualization task. To validate the efficacy of SEMAFORE, our comprehensive experiments on a number of real-life text datasets of news articles and Web pages show that the proposed methods outperform the state-of-the-art baselines on objective evaluation metrics.

Semantic visualization with neighborhood graph regularization

Visualization of high-dimensional data, such as text documents, is useful to map out the similarities among various data points. In the high-dimensional space, documents are commonly represented as...

Altered Expression of Functional Proteins in CD4 Regulatory T Cells during Therapy with Idelalisib

Introduction: Idelalisib, a specific small molecule inhibitor of the delta isoform of PI3 kinase (PI3Kd), has recently been FDA approved for the treatment of relapsed CLL. Expression of PI3Kd is predominantly restricted to leukocytes, and inhibition of PI3Kd promotes apoptosis of neoplastic CLL cells. Patients who receive idelalisib often experience toxicities that appear to be immune mediated. In an ongoing phase II study of idelalisib-ofatumumab at DFCI, 16 of 21 patients experienced grade 3 or greater autoimmune toxicities (14 transaminitis, 2 enteritis/colitis and 3 presumed drug-induced pneumonitis).Methods: To identify potential causes of autoimmunity in patients receiving idelalisib, we used single cell mass cytometry (CyTOF) with a panel of 26 surface membrane and 9 intracellular markers to provide a comprehensive phenotypic and functional analysis of all peripheral blood lymphocytes. Blood samples were obtained at baseline prior to therapy and at onset of autoimmune symptoms, which occurred at a median of 28 days from start of therapy. Results in 5 severely affected patients were compared with 3 patients who did not develop autoimmune toxicities and 15 healthy donors. The second sample from unaffected patients was matched in time to the affected patient samples.Results: Within CD3 T cells, idelalisib therapy led to an increased percentage of CD4 T cells and decreased percentage of CD8 T cells. Within CD4 T cells, CD25high Foxp3+ regulatory T cells (Treg) were reduced, resulting in a marked increase of the ratio of conventional CD4 T (Tcon)/Treg and increased CD8/Treg ratio. These findings were similar in patients who developed autoimmune toxicities and those who did not. Analysis of effector memory (EM; 45RA- CCR7-), central memory (CM; 45RA- CCR7+), terminal effector (TEMRA; 45RA+ CCR7-) and naive (45RA+ CCR7+) subsets within Treg, Tcon and CD8 T cells was comparable before and after idelalisib therapy and with healthy individual controls. To evaluate the heterogeneity of each population we utilized viSNE, which allows visualization of complex high-dimensional cytometry data on a two-dimensional plot. Treg in patients who developed autoimmune phenomena were compared with Treg from patients who did not. In patients who developed autoimmunity, baseline Treg had higher expression of PD-1 and lower expression of functional markers, such as GITR, Tbet, CXCR3, PDL-1, granzyme-β and TIM-3. In some cases, expression of these functional markers declined further with idelalisib therapy (Figure 1). Cytotoxic T lymphocyte antigen 4 (CTLA4) is constitutively expressed by Treg, and its deficiency has been shown to limit suppressive function of Treg. Inducible co-stimulator (ICOS) also mediates the suppressive functions of CD4 Treg by regulating cell surface expression of CTLA-4. Interestingly, expression of CTLA-4 and ICOS did not change during idelalisib therapy (Figure 1). Susceptibility to apoptosis was monitored by expression of pro-survival Bcl-2 and pro-apoptotic CD95 (FAS). Treg from patients developing autoimmunity expressed lower levels of Bcl-2 and higher levels of CD95 compared to those who did not. Conversely, Treg of non-affected patients retained higher expression of functional markers and became more active, assessed by increased expression of HLA-DR. The use of CyTOF combined with our comprehensive panel allowed us to delineate up to 13 distinct Treg sub-populations among healthy controls. In contrast, CLL patients often lacked such broad Treg heterogeneity. Expression of functional markers by Tcon and CD8 T cells from patients who would develop side effects, were also slightly lower, suggesting that autoimmune toxicities were not due to expansion or activation of CD8 or Tcon effector cells but due to the defect in both number and function of Tregs, which became insufficient to maintain the immune tolerance.Conclusion: These studies allowed us to identify defects in both Treg number and function, which may be responsible for the autoimmune toxicities in patients receiving idelalisib therapy. Further studies may allow the development of a reliable predictor of these toxicities by evaluation of the Treg markers prior to drug exposure which could be used to guide therapy. Moreover, the identification of idelalisib as a Treg inhibitor suggests that it could be used as an immunomodulatory agent to target the Treg pathway, in patients with solid tumors. [Display omitted] DisclosuresNo relevant conflicts of interest to declare.

An enhanced data visualization method for diesel engine malfunction classification using multi-sensor signals.

The various multi-sensor signal features from a diesel engine constitute a complex high-dimensional dataset. The non-linear dimensionality reduction method, t-distributed stochastic neighbor embedding (t-SNE), provides an effective way to implement data visualization for complex high-dimensional data. However, irrelevant features can deteriorate the performance of data visualization, and thus, should be eliminated a priori. This paper proposes a feature subset score based t-SNE (FSS-t-SNE) data visualization method to deal with the high-dimensional data that are collected from multi-sensor signals. In this method, the optimal feature subset is constructed by a feature subset score criterion. Then the high-dimensional data are visualized in 2-dimension space. According to the UCI dataset test, FSS-t-SNE can effectively improve the classification accuracy. An experiment was performed with a large power marine diesel engine to validate the proposed method for diesel engine malfunction classification. Multi-sensor signals were collected by a cylinder vibration sensor and a cylinder pressure sensor. Compared with other conventional data visualization methods, the proposed method shows good visualization performance and high classification accuracy in multi-malfunction classification of a diesel engine.

Constrained Self Organizing Maps for Data Clusters Visualization

High dimensional data visualization is one of the main tasks in the field of data mining and pattern recognition. The self organizing maps (SOM) is one of the topology visualizing tool that contains a set of neurons that gradually adapt to input data space by competitive learning and form clusters. The topology preservation of the SOM strongly depends on the learning process. Due to this limitation one cannot guarantee the convergence of the SOM in data sets with clusters of arbitrary shape. In this paper, we introduce Constrained SOM (CSOM), the new version of the SOM by modifying the learning algorithm. The idea is to introduce an adaptive constraint parameter to the learning process to improve the topology preservation and mapping quality of the basic SOM. The computational complexity of the CSOM is less than those with the SOM. The proposed algorithm is compared with similar topology preservation algorithms and the numerical results on eight small to large real-world data sets demonstrate the efficiency of the proposed algorithm.

RFM 기반 SOM을 이용한 매장관리 전략 도출

RFM 기반 SOM을 이용한 매장관리 전략 도출

Critical data points-based unsupervised linear dimension reduction technology for science data

Recent advances in machine learning and data mining have led to powerful methods for the analysis and visualization of high dimensional data. This paper proposes an unsupervised linear dimension re...

MLN0128, a Second-Generation mTOR Kinase Inhibitor, Disrupts Survival Signaling and Triggers Apoptosis in AML

Abnormal mTOR activation leads to enhanced survival signaling in acute myeloid leukemia (AML) cells. The active site mTOR kinase inhibitors (TOR-KI) represent a promising new approach to target the PI3K/AKT/mTOR pathway. In a previous study, we used a synthetic TOR-KI PP242 to demonstrate that disrupting AKT/mTOR signaling with this class of compounds can effectively target leukemic cells under conditions mimicking bone marrow microenvironment[1]. In the current study, we focused on the inhibitory mechanisms in rare and heterogeneous AML stem cells and on the cellular antagonistic responses after mTORC1/2 blockage using MLN0128, a synthetic TOR-KI currently in clinical development. We found that MLN0128 inhibited cell proliferation and induced apoptosis in AML cells by attenuating mTOR complex 1 and 2 (mTORC1/2) activities, and showed higher anti-leukemia potency than PP242. Using high-throughput time-of-flight mass cytometry (CyTOF)[2], combined with advanced computational analytic software Spanning-Tree Progression of Density-Normalized Events (SPADE)[3] and visualization t-Distribution Stochastic Neighbor Embedding (viSNE)[4], we identified AML subsets co-expressing multiple stem cell markers with high AKT/mTOR activity. MLN0128 selectively targets mTORC1/2 downstream signaling in these cells with minimal effects on other pathways, and lacks the effect on non- leukemic stem cells (LSC). The inhibitory effect of MLN0128 on LSC is positively correlated to that of a number of co-expressing stem cell molecules, and to the baseline expression of and the degree of SCF activation on AKT/mTOR signaling. Using the reverse phase protein array (RPPA) technique, we studied expression and phosphorylation changes in 151 proteins from 24 primary AML samples in response to MLN0128, and revealed several pro-survival pathways may protect from MLN0128-induced cellular stress. We identified four druggable targets that were significantly triggered by MLN0128 inhibition: PARP (that can be targeted by PARP inhibitor olaparib), HDAC3 (vorinostat), p-SRC (dasatinib), and STAT5 (ruxolitinib). Blocking AKT/mTOR signaling and these pro-survival cascades triggered complex responses in AML cell lines and primary AML samples, resulting in an overall increase in apoptosis of AML cells compared to single target inhibition. Among the four combinations, MLN0128 and vorinostat elicited the best responses. In summary, our findings indicate that blocking AKT/mTOR signaling with the mTORC1/2 inhibitor MLN0128 causes growth inhibition and apoptosis in AML cells. High-throughput CyTOF technology revealed that MLN0128 selectively targets primary AML stem cells with high AKT/mTOR activity. RPPA mapping multiple intracellular signaling pathways in primary AML cells in response to mTORC1/2 inhibition allowed us to identify compensatory activation of actionable pro-survival pathways. Combinatorial multi-targeted approaches demonstrated various responses to MLN0128 in AML samples, suggesting that personalized treatment selections will be needed in individual AML cases. Finally, utilization of techniques such as CyTOF or RPPA may assist in profiling intra-cellular and inter-patient variability of responses to targeted agents which likely stem from multiple genetic and epigenetic events within individual tumor, and serve as tools for selection of individualized targeted therapies in AML and other malignancies.Reference1. Zeng, Z.H., et al., Targeting of mTORC1/2 by the mTOR kinase inhibitor PP242 induces apoptosis in AML cells under conditions mimicking the bone marrow microenvironment. Blood, 2012. 120(13): p. 2679-2689.2. Bendall, S.C., et al., Single-Cell Mass Cytometry of Differential Immune and Drug Responses Across a Human Hematopoietic Continuum. Science, 2011. 332(6030): p. 687-696.3. Qiu, P., et al., Extracting a cellular hierarchy from high-dimensional cytometry data with SPADE. Nature Biotechnology, 2011. 29(10): p. 886-U181.4. Amir el, A.D., et al., viSNE enables visualization of high dimensional single-cell data and reveals phenotypic heterogeneity of leukemia. Nat Biotechnol, 2013. 31(6): p. 545-52. DisclosuresNo relevant conflicts of interest to declare.

Doubly supervised embedding based on class labels and intrinsic clusters for high-dimensional data visualization

Visualization of data can assist decision-making processes by presenting the underlying information in a perceptible manner. Many dimension reduction techniques have been proposed to generate faithful visualization snapshots given high-dimensional data. When class labels associated with the data are already provided, supervised dimension reduction methods, which utilize such pre-given label information as well as the data, have been effective in revealing the overall structure of data with respect to their pre-given class labels. However, the main principle of most of these supervised methods has been to enhance class separability, which generally leads to significant distortion of original relationships. To compensate for such distortion, we propose a novel doubly supervised dimension reduction approach that highlights both natural groupings conforming to original relationships and classes determined by pre-given labels. Our method imposes minimal supervision on the pre-given class information depending on their original distributions while imposing additional supervision on natural groupings to better preserve them in reduced feature space. Specifically, we apply the notion of doubly supervised dimension reduction to a state-of-the-art method called t-distributed stochastic neighbor embedding and present a new formulation and an algorithm. By performing both quantitative and qualitative analyses, we demonstrate the effectiveness of our method using various visualization examples on real-world data. Our results show that, compared to other existing methods, the proposed method better preserves the original high-dimensional relationships while simultaneously maintaining class separability and preserving cluster structures. In addition, due to the characteristics of preserving natural groupings, the visualization results generated by our method reveal interesting sub-groups that cohesively preserve the original relationships in the data.

Scatterplot layout for high-dimensional data visualization

Multi-dimensional data visualization is an important research topic that has been receiving increasing attention. Several techniques that apply scatterplot matrices have been proposed to represent multi-dimensional data as a collection of two-dimensional data visualization spaces. Typically, when using the scatterplot-based approach it is easier to understand relations between particular pairs of dimensions, but it often requires too large display spaces to display all possible scatterplots. This paper presents a technique to display meaningful sets of scatterplots generated from high-dimensional datasets. Our technique first evaluates all possible scatterplots generated from high-dimensional datasets, and selects meaningful sets. It then calculates the similarity between arbitrary pairs of the selected scatterplots, and places relevant scatterplots closer together in the display space while they never overlap each other. This design policy makes users easier to visually compare relevant sets of scatterplots. This paper presents algorithms to place the scatterplots by the combination of ideal position calculation and rectangle packing algorithms, and two examples demonstrating the effectiveness of the presented technique.

Parametric nonlinear dimensionality reduction using kernel t-SNE

Abstract Novel non-parametric dimensionality reduction techniques such as t-distributed stochastic neighbor embedding (t-SNE) lead to a powerful and flexible visualization of high-dimensional data. One drawback of non-parametric techniques is their lack of an explicit out-of-sample extension. In this contribution, we propose an efficient extension of t-SNE to a parametric framework, kernel t-SNE, which preserves the flexibility of basic t-SNE, but enables explicit out-of-sample extensions. We test the ability of kernel t-SNE in comparison to standard t-SNE for benchmark data sets, in particular addressing the generalization ability of the mapping for novel data. In the context of large data sets, this procedure enables us to train a mapping for a fixed size subset only, mapping all data afterwards in linear time. We demonstrate that this technique yields satisfactory results also for large data sets provided missing information due to the small size of the subset is accounted for by auxiliary information such as class labels, which can be integrated into kernel t-SNE based on the Fisher information.

Visualization and clustering of high-dimensional transcriptome data using GATE.

The potential gains from advances in high-throughput experimental molecular biology techniques are commonly not fully realized since these techniques often produce more data than can be easily organized and visualized. To address these problems, GATE (Grid-Analysis of Time-Series Expression) was developed. GATE is an integrated software platform for the analysis and visualization of high-dimensional time-series datasets, which allows flexible interrogation of time-series data against a wide range of databases of prior knowledge, thus linking observed molecular dynamics to potential genetic, epigenetic, and signaling mechanisms responsible for observed dynamics. This article provides a brief guide to using GATE effectively.

Accelerating t-SNE using tree-based algorithms

The paper investigates the acceleration of t-SNE--an embedding technique that is commonly used for the visualization of high-dimensional data in scatter plots--using two tree-based algorithms. In particular, the paper develops variants of the Barnes-Hut algorithm and of the dual-tree algorithm that approximate the gradient used for learning t-SNE embeddings in O(N log N). Our experiments show that the resulting algorithms substantially accelerate t-SNE, and that they make it possible to learn embeddings of data sets with millions of objects. Somewhat counterintuitively, the Barnes-Hut variant of t-SNE appears to outperform the dual-tree variant.

Design and outcomes of an acoustic data visualization seminar

Recently, the Department of Media Technology at Aalto University offered a seminar entitled Applied Data Analysis and Visualization. The course used spatial impulse response measurements from concert halls as the context to explore high-dimensional data visualization methods. Students were encouraged to represent source and receiver positions, spatial aspects, and temporal development of sound fields, frequency characteristics, and comparisons between halls, using animations and interactive graphics. The primary learning objectives were for the students to translate their skills across disciplines and gain a working understanding of high-dimensional data visualization techniques. Accompanying files present examples of student-generated, animated and interactive visualizations.

SketchPadN-D: WYDIWYG Sculpting and Editing in High-Dimensional Space

High-dimensional data visualization has been attracting much attention. To fully test related software and algorithms, researchers require a diverse pool of data with known and desired features. Test data do not always provide this, or only partially. Here we propose the paradigm WYDIWYGS (What You Draw Is What You Get). Its embodiment, SketchPadND, is a tool that allows users to generate high-dimensional data in the same interface they also use for visualization. This provides for an immersive and direct data generation activity, and furthermore it also enables users to interactively edit and clean existing high-dimensional data from possible artifacts. SketchPadND offers two visualization paradigms, one based on parallel coordinates and the other based on a relatively new framework using an N-D polygon to navigate in high-dimensional space. The first interface allows users to draw arbitrary profiles of probability density functions along each dimension axis and sketch shapes for data density and connections between adjacent dimensions. The second interface embraces the idea of sculpting. Users can carve data at arbitrary orientations and refine them wherever necessary. This guarantees that the data generated is truly high-dimensional. We demonstrate our tool's usefulness in real data visualization scenarios.