T-stochastic Neighbor Embedding Research Articles

Embedding Functional Brain Networks in Low Dimensional Spaces Using Manifold Learning Techniques.

Background: fMRI data is inherently high-dimensional and difficult to visualize. A recent trend has been to find spaces of lower dimensionality where functional brain networks can be projected onto manifolds as individual data points, leading to new ways to analyze and interpret the data. Here, we investigate the potential of two powerful non-linear manifold learning techniques for functional brain networks representation: (1) T-stochastic neighbor embedding (t-SNE) and (2) Uniform Manifold Approximation Projection (UMAP) a recent breakthrough in manifold learning.Methods: fMRI data from the Human Connectome Project (HCP) and an independent study of aging were used to generate functional brain networks. We used fMRI data collected during resting state data and during a working memory task. The relative performance of t-SNE and UMAP were investigated by projecting the networks from each study onto 2D manifolds. The levels of discrimination between different tasks and the preservation of the topology were evaluated using different metrics.Results: Both methods effectively discriminated the resting state from the memory task in the embedding space. UMAP discriminated with a higher classification accuracy. However, t-SNE appeared to better preserve the topology of the high-dimensional space. When networks from the HCP and aging studies were combined, the resting state and memory networks in general aligned correctly.Discussion: Our results suggest that UMAP, a more recent development in manifold learning, is an excellent tool to visualize functional brain networks. Despite dramatic differences in data collection and protocols, networks from different studies aligned correctly in the embedding space.

Single-Cell RNA Sequencing Unveils the Hallmarks of Populations at Different Stages of HTLV-1 Infection

BackgroundAdult T-cell leukemia-lymphoma (ATL) is an aggressive mature T-cell neoplasm caused by human T -cell leukemia virus type 1 (HTLV-1). Up to 5% of infected individuals develop to ATL. HTLV-1 preferentially infects CD4 + T cells, and stimulates cell proliferation and prevents cell death by apoptosis. The viral oncogene-encoded proteins, Tax and HBZ, play important roles in viral infection and cell immortalization. However, the host factor of developing from carrier to patient is not clear.ResultsTo characterize the heterogeneity of ATL patients, we performed single-cell RNA-sequencing (10x Genomics) analysis on single cell suspensions isolated from PBMCs of nine samples, including three ATL patients, three HTLV-1 asymptomatic carriers as well as three healthy donors (HD). We acquired 82977 high-quality cells with a median of 1718 genes detected per cell. Unsupervised clustering using Seurat followed by visualization in t-Stochastic Neighbor Embedding (t-SNE) identified 29 distinct cell clusters (Figure 1A). The single cell profiling of ATL patients were significantly different from that of carriers and healthy donors, while the latter two had little difference (Figure 1B). Based on singleR packages and marker genes of each cluster, 4 major cell populations (T cells, NK cells, B cells and myeloid cells) and other rare cell types were annotated, such as erythrocyte cluster and eosinophils cluster.We observed an enrichment of CD4 + T cell from patients in 4 cluster (Figure 1C), which proportion of cells was higher than that of carriers and healthy donors. According to cell type annotation, cells from cluster 11 were CD4 + CD25 + Foxp3 + Treg cells. Based on the increasing proportion of cluster 11 in healthy people, carriers and patients, without significant statistical differences, we assumed that Foxp3 + Treg cells were involved in the evolution of ATL tumor cells. That was identical with published literatures.To investigate the differences between tumor and normal CD4 + T cell, the gene expression was compared among 7 clusters of CD4 + T cell from three groups. Using a threshold of p value < 0.05 and | fold change| >2. Through integrated analysis, we identified 26 commonly upregulated genes (gene expression level: patients > carriers > HD) and 9 downregulated genes (gene expression level: patients < carriers < HD. To further analyze the biological function of the common DEGs, gene ontology (GO) analysis showed that these genes could be mainly categorized into plasma membrane and protein binding.Subsequently, we validated the mRNA expression level of upregulated common DEGs among three groups by qRT-PCR. The isolated CD4 + T cell using CD4 microbeads of a total of 6 patients, 3 carriers and 9 normal specimens were included. The result showed that the mRNA expression levels of gene CADM1 and RGS13 in patients were higher than those in carriers and healthy donors, although there was no statistical difference between patients and carriers, and the expression levels of carriers tended to be higher than those in normal people (Figure 1D and E). Previously, CADM1 has been revealed to be highly expressed in HTLV‐1‐infected CD4 + T cells. Our study confirmed this result by single-cell profiling. RGS13, a member of the regulators of G protein signaling (RGS) family, participates in cellular communication. The role of RGS13 in ATL needs to be investigated.ConclusionsThis study is the first time to analyze the single-cell RNA level of ATL patients, HTLV-1 virus carriers and normal people. The peripheral blood cell composition of the patient is significantly different from that of the carriers and healthy donors, while it is similar between carriers and normal people. CD4 + T cells are the main cell population of patients. The proportion of CD4 + CD25 + Foxp3 + Treg cells increased gradually in healthy people, carriers and patients. DEGs analysis showed that CADM1 and RGS13 were highly expressed in CD4 + T cells of patients, followed by carriers, validated by 18 clinical samples. However, the molecular mechanism of RGS13 in the process from HTLV-1 infection to ATL needs to be further studied. [Display omitted] DisclosuresHu: Astellas Pharma, Inc.: Research Funding.

Detecting Multidimensional Pareto Fronts During the Optimization of a Return Channel Within a Centrifugal Compressor

Abstract We explain the advantages of using big-data methods for evaluating numerical optimization. The investigation focuses on the performance potential of three-dimensional return channel vanes under realistic manufacturing constraints. Based on an analysis of an optimization database, this paper presents a systematic approach for analysis and design guidelines for three-dimensional return channel vanes. To this end, a validated numerical setup was developed on the basis of experimental investigations, followed by a numerical optimization using a genetic algorithm and an artificial neural network. The optimization database was analyzed with a dimension reduction method called t-stochastic neighbor embedding. This method enabled linking geometric design features with physical correlations and, finally, with the objective functions of the optimization. With the help of the detected correlations within the database, it has been possible to work out a method for deciding on the selection of a design on the Pareto front and to draw new relevant conclusions. The systematic use of big-data methods proposed enables a more penetrating insight to be gained into numerical optimization, which is more general and relevant than those gained by simply comparing a single optimized design with a reference design. An analysis of the Pareto front reveals that 0.6% efficiency can be exchanged for 20% more homogeneous outflow.

Identification of a Hematopoietic Stem Cell Fraction That Retains Normal Function Following Chronic Inflammation

Chronic inflammation has been suggested to impair hematopoietic stem cell (HSC) function, resulting in decreased self-renewal and myeloid overproduction at the expense of other lineages. We previously identified the pro-inflammatory cytokine interleukin (IL)-1 as a key activator of precocious HSC differentiation that simultaneously impairs HSC self-renewal. However, recent identification of multiple HSC subsets with distinct functional and molecular characteristics calls into question whether the HSC compartment is uniformly affected by chronic inflammatory signals. Moreover, a growing body of work suggests that traditional markers in the mouse that define HSC potential under homeostatic conditions, such as Sca-1 and CD34, undergo expression changes that limit their utility in prospective identification of functional HSCs during non-homeostatic conditions such as inflammation. Here we use injection of IL-1β to model chronic inflammation in the mouse, and demonstrate that the rare EPCR+CD34- fraction contained within the phenotypic Lin-cKit+Sca-1+Flk2-CD48-CD150+ (SLAM) HSC compartment is uniquely able to maintain its function despite chronic IL-1 exposure.To assess the impact of chronic inflammation on HSC compartment dynamics, we injected wild-type (WT) C57BL/6 mice with recombinant murine IL-1β for 20 days, followed by analysis of HSC populations using multi-parametric flow-cytometry. Strikingly, t-Stochastic Neighbor Embedding (tSNE) analysis of our flow cytometry data identified changes in the relative abundance of several novel HSC sub-populations following IL-1 exposure, including a sharp decrease in the number of CD34- cells also expressing the long-term HSC (LT-HSC) marker EPCR (CD201). Consistent with the characteristics of a long-term HSC, CD34-EPCR+ HSCs exhibited slow growth and differentiation kinetics in liquid culture, and elevated colony forming potential in serial re-plating. Moreover, molecular analyses demonstrated EPCR+CD34- HSCs are uniquely enriched for expression of LT-HSC genes including Hoxb5, Fgd5 and Cpt1a . Nonetheless, EPCR+CD34- cells express Il1r1 and underwent accelerated myeloid differentiation in the presence of IL-1, suggesting they are directly responsive to this cytokine.To determine the impact of sustained IL-1 exposure on the blood-forming capacity of EPCR+CD34- HSCs in vivo, we assessed their growth and potential ex vivo following treatment of mice with PBS or IL-1β for 20 days. Notably, prior exposure to IL-1 did not significantly alter the growth or short-term potential of EPCR+CD34- HSCs based on in vitro assays. Strikingly, ongoing transplantation experiments of IL-1-exposed EPCR+CD34- HSCs versus total SLAM HSCs into irradiated recipient mice have revealed that while the total SLAM HSC compartment exhibits immediate impairment of reconstitution capacity, EPCR+CD34- HSCs exhibit no functional impairment, suggesting IL-1 signaling may not directly impact their self-renewal capacity. We are currently examining the molecular and epigenetic features of these cells to understand how they retain their self-renewal capacity in the face of inflammation, and assessing whether this compartment retains its functionality in a range of inflammatory disease contexts.Overall, our findings have uncovered a distinct fraction within the phenotypic SLAM HSC that retains essentially normal function despite exposure to chronic inflammation. This supports the notion that HSCs respond to inflammation in a heterogeneous manner and that a functional LT-HSC compartment is present even during chronic inflammatory contexts. These investigations stand to provide critical insights that will increase our understanding of the impact of inflammation on long-term blood system function in the context of chronic inflammatory disease, bone marrow transplant and hematological malignancy. DisclosuresNo relevant conflicts of interest to declare.

Therapeutic potential of targeting Tfr/Tfh cell balance by low-dose-IL-2 in active SLE: a post hoc analysis from a double-blind RCT study

ObjectiveTo investigate the regulation of T follicular regulatory (Tfr) and T follicular (Tfh) cell subtypes by low-dose IL-2 in systemic lupus erythematosus (SLE) in a randomized, double-blind, placebo-controlled clinical trial.MethodsA post hoc analysis was performed in a randomized cohort of SLE patients (n=60) receiving low-dose IL-2 therapy (n=30) or placebo (n=30), along with the standard of care treatment. The primary endpoint was the attainment of SLE responder index-4 (SRI-4) at week 12 in the trial. Twenty-three healthy controls were enrolled for T cell subset detection at the same time as the trial. The t-stochastic neighbor embedding (tSNE) analysis of CD4 T subsets based on immune cells flow cytometry markers was performed to distinguish Tfh, Tfh1, Tfh2, Tfh17, and Tfr cell subsets.ResultsCompared with HC, the frequency of Tfr (CXCR5+PD-1low Treg and CXCR5+PD-1high Treg) cells was significantly reduced, while the pro-inflammatory Tfh cells were increased in patients with SLE. The imbalanced Tfh cell was associated with several pathogenic factors (anti-dsDNA antibodies (r=0.309, P=0.027) and serum IL-17 (r=0.328, P=0.021)) and SLE Disease Activity Index (SLEDAI) score (r=0.273, P=0.052). Decreased CXCR5+PD-1low Treg/Tfh and CXCR5+PD-1low Treg/Tfh17 were both associated with increased immunoglobulin M (IgM) (r=−0.448, P=0.002 and r=−0.336, P=0.024, respectively). Efficacy of low-dose IL-2 therapy was associated with a restored Tfr/Tfh cell balance.ConclusionThese data support the hypothesis that promotion of Tfr is associated with decreased disease activities and that low-dose IL-2 therapy can recover Tfr/Tfh immune balance.Trial registration numberClinicalTrials.gov Registries (NCT02465580).

Capturing the Development of Internal Representations in a High-Performing Deep Network for Sleep Stage Classification

Deep neural networks can provide accurate automated classification of human sleep signals into sleep stages that enables more effective diagnosis and treatment of sleep disorders. We develop a deep convolutional neural network (CNN) that attains state-of-the-art sleep stage classification performance on input data consisting of human sleep EEG and EOG signals. Nested cross-validation is used for optimal model selection and reliable estimation of out-of-sample classification performance. The resulting network attains a classification accuracy of 84.50 pm 0.13%; its performance exceeds human expert inter-scorer agreement, even on single-channel EEG input data, therefore providing more objective and consistent labeling than human experts demonstrate as a group. We focus on analyzing the learned internal data representations of our network, with the aim of understanding the development of class differentiation ability across the layers of processing units, as a function of layer depth. We approach this problem visually, using t-Stochastic Neighbor Embedding (t-SNE), and propose a pooling variant of Centered Kernel Alignment (CKA) that provides an objective quantitative measure of the development of sleep stage specialization and differentiation with layer depth. The results reveal a monotonic progression of both of these sleep stage modeling abilities as layer depth increases.

Investigation of Lipid Metabolism in Dynamic Progression of Coronary Artery Atherosclerosis of Humans by Time-of-Flight Secondary Ion Mass Spectrometry.

Alterations in lipid metabolites in coronary artery tissues are phenotypic changes in the progression of atherosclerosis (AS). A full picture of the spatiotemporal distribution of lipid metabolites in coronary AS is needed for a deeper understanding of its pathology and the identification of potential biomarkers of disease progression. In this work, the changes in species, quantity, and distribution of lipid metabolites at different stages of AS, which were standardized by the disease areas, were analyzed through the high spatial resolution- and high sensitivity-time-of-flight secondary ion mass spectrometry (ToF-SIMS) under delayed extraction mode. Based on high lateral resolution imaging, we further analyzed the ToF-SIMS data extracted from the subregions of AS lesion tissues at different disease progression stages by semiquantitative comparison, clustering analysis (t-stochastic neighbor embedding and HCA), and KEGG enrichment. Thus, a much-detailed description of lipids' features in coronary AS was achieved. We constructed a ToF-SIMS mass spectrometry database of coronary AS lipids. 40 specific lipid metabolites with distinctive patterns between different pathological stages were obtained. Chemical imaging unveiled further details regarding the spatial distribution of lipids. Moreover, linoleic acid and arachidonic acid metabolic pathway were predicted to be critical in AS progression.

On the ability to exploit signal fluctuations in pseudocontinuous arterial spin labeling for inferring the major flow territories from a traditional perfusion scan

In arterial spin labeling (ASL) a magnetic label is applied to the flowing blood in feeding arteries allowing depiction of cerebral perfusion maps. The labeling efficiency depends, however, on blood velocity and local field inhomogeneities and is, therefore, not constant over time. In this work, we investigate the ability of statistical methods used in functional connectivity research to infer flow territory information from traditional pseudo-continuous ASL (pCASL) scans by exploiting artery-specific signal fluctuations. By applying an additional gradient during labeling the minimum amount of signal fluctuation that allows discrimination of the main flow territories is determined. The following three approaches were tested for their performance on inferring the large vessel flow territories of the brain: a general linear model (GLM), an independent component analysis (ICA) and t-stochastic neighbor embedding. Furthermore, to investigate the effect of large vessel pathology, standard ASL scans of three patients with a unilateral stenosis (>70%) of one of the internal carotid arteries were retrospectively analyzed using ICA and t-SNE. Our results suggest that the amount of natural-occurring variation in labeling efficiency is insufficient to determine large vessel flow territories. When applying additional vessel-encoded gradients these methods are able to distinguish flow territories from one another, but this would result in approximately 8.5% lower perfusion signal and thus also a reduction in SNR of the same magnitude.

Hybrid rail track quality analysis using nonlinear dimension reduction technique with machine learning

Track geometry parameters from rail track inspection are regulated within unique safety limits for different track classes. This study focuses on developing an index that combines safety and track quality because of the inefficiency of corrective maintenance activities between routine maintenance cycles when federal geometry limits are violated. This combination is achievable by summarizing multivariate track geometry parameters as an improvement to previous linear approaches to address the problem of inefficient track geometry maintenance programs. The use of nonlinear dimension reduction (T-stochastic neighbor embedding (T-SNE)) for hybrid track quality index development and the influence of time-based parameters on track quality is evaluated in this study. The results show that the probability of geometry defects is correlated with principal components, but T-SNE had the best prediction on train-test splits despite its poor performance on a blind validation set. The absence of an observable correlation between the track geometry and acceleration data requires further investigation.

Behavior analysis method for indoor environment based on app usage mining

With the recent development of information and communication technologies, the utilization areas of spatial information are increasing rapidly while merging with various industries and technologies. This study introduces a system that can analyze the space utilization of users at low cost in indoor environment using simple smartphone app logs. We collect and process important information from mobile app logs and Google app server and generate a high-dimensional dataset required to analyze user behaviors. In addition, user behaviors are classified and clustered by applying a VGG classifier and a clustering algorithm based on t-stochastic neighbor embedding (t-SNE). Our system can easily acquire a large amount of data required for deep learning network learning without additional sensors for spatial analysis and enhance the accuracy of network classification and cluster through these data. Our methodology can assist spatial analysis for indoor environments in which people are living and help reduce the cost of space utilization feedback from users.

Review of Dimension Reduction Methods

Purpose: This study sought to review the characteristics, strengths, weaknesses variants, applications areas and data types applied on the various Dimension Reduction techniques. Methodology: The most commonly used databases employed to search for the papers were ScienceDirect, Scopus, Google Scholar, IEEE Xplore and Mendeley. An integrative review was used for the study where 341 papers were reviewed. Results: The linear techniques considered were Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), Singular Value Decomposition (SVD), Latent Semantic Analysis (LSA), Locality Preserving Projections (LPP), Independent Component Analysis (ICA) and Project Pursuit (PP). The non-linear techniques which were developed to work with applications that have complex non-linear structures considered were Kernel Principal Component Analysis (KPCA), Multi-dimensional Scaling (MDS), Isomap, Locally Linear Embedding (LLE), Self-Organizing Map (SOM), Latent Vector Quantization (LVQ), t-Stochastic neighbor embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP). DR techniques can further be categorized into supervised, unsupervised and more recently semi-supervised learning methods. The supervised versions are the LDA and LVQ. All the other techniques are unsupervised. Supervised variants of PCA, LPP, KPCA and MDS have been developed. Supervised and semi-supervised variants of PP and t-SNE have also been developed and a semi supervised version of the LDA has been developed. Conclusion: The various application areas, strengths, weaknesses and variants of the DR techniques were explored. The different data types that have been applied on the various DR techniques were also explored.

Joint Characterization of Multiscale Information in High Dimensional Data

High dimensional data can contain multiple scales of variance. Analysis tools that preferentially operate at one scale can be ineffective at capturing all the information present in this cross-scale complexity. We propose a multiscale joint characterization approach designed to exploit synergies between global and local approaches to dimensionality reduction. We illustrate this approach using Principal Components Analysis (PCA) to characterize global variance structure and t-stochastic neighbor embedding (t-sne) to characterize local variance structure. Using both synthetic images and real-world imaging spectroscopy data, we show that joint characterization is capable of detecting and isolating signals which are not evident from either PCA or t-sne alone. Broadly, t-sne is effective at rendering a randomly oriented low-dimensional map of local clusters, and PCA renders this map interpretable by providing global, physically meaningful structure. This approach is illustrated using imaging spectroscopy data, and may prove particularly useful for other geospatial data given robust local variance structure due to spatial autocorrelation and physical interpretability of global variance structure due to spectral properties of Earth surface materials. However, the fundamental premise could easily be extended to other high dimensional datasets, including image time series and non-image data.

Blinatumomab Consolidation Post Autologous Hematopoietic Stem Cell Transplantation in Patients with Diffuse Large B Cell Lymphoma

Introduction: Autologous hematopoietic stem cell transplantation (auto-HCT) is the standard treatment for patients with chemo-sensitive relapsed/refractory diffuse large B cell lymphoma (DLBCL). However, post-auto-HCT outcomes are still poor in this population, with 5-year progression free survival (PFS) of 40%. We hypothesize that in patients with DLBCL, blinatumomab consolidation post auto-HCT will eradicate remaining tumor cells, leading to decreased relapse and increased overall survival. Therefore we conducted a pilot study to test blinatumomab as consolidation therapy post auto-HCT for patients with DLBCL. Methods: Adult patients with chemosensitive DLBCL or transformed FL who underwent auto-HCT were included. All patients received one cycle of blinatumomab consolidation starting 42 days post auto-HCT (9 mcg daily as continuous infusion for 7 days, followed by 28 mcg daily for 21 days). Response evaluation was done at day 100 post auto-HCT. Minimal residual disease (MRD) was quantified by immunoglobulin high-throughput sequencing (Ig-HTS) of plasma cell-free DNA on days 42 post auto-HCT (pre-blinatumomab) and on day 100 post auto-HSCT (one month post completion of blinatumomab). Immunophenotyping of T cells in cryopreserved peripheral blood mononuclear cells collected on day 42 (pre-blinatumomab), day 56 (midpoint of blinatumomab treatment cycle), and day 100 (1 month post blinatumomab) was performed using 18-color flow cytometry panels for extracellular and intracellular antigens. Results: As of August 2020, ten patients have been treated with at least 100 days follow up. Patient characteristics and outcomes are summarized in Table 1. Three out of 10 patients (30%) were in partial remission (PR) as determined by CT or PET/CT imaging before auto-HCT. All subjects completed the planned cycle of blinatumomab consolidation. Blinatumomab was well tolerated. Two patients developed grade 1 CRS, with no grade 2 and higher CRS. Immune effector cell-associated neurotoxicity syndrome (ICANS) was not observed. Six patients developed transient tremor (four grade 1, one grade 2, and one grade 3). One subjects developed BCNU pneumonitis and CMV viremia that resolved with steroid and ganciclovir. One hundred days post auto-HCT (one month post blinatumomab consolidation) 10/10 (100%) of patient were in complete remission (CR) as determined by both MRD testing and by CT or PET/CT imaging. Plasma cell free based MRD was positive on day 42 (post auto-HCT and pre-blinatumomab), in two out of ten patients (20%). These two patients achieved MRD negative status after receiving blinatumomab consolidation. With median follow up of 14.5 months (range: 7-34 months), all 10 patients are alive and 6/10 remain in remission.. Interestingly, the 4 patients with disease relapse had lower CD8/CD4 T cell ratio before starting blinatumomab compared with patients who remained in remission (Figure 1A). However, there were no significant differences in the distribution of the major T cell subtypes (naïve, memory, effector and Treg), and expression of markers of T cell activation, proliferation, or exhaustion (Figure 1B-1E). High dimensional analysis with t-stochastic neighbor embedding (tSNE) revealed a cluster of CD8+ and CD4+ T cells characterized by high expression of granzyme B (GB) and perforin that was present in the DLBCL patients before and after blinatumomab treatment but not in a healthy untreated control (Figure 1F-1H). Although further analysis of healthy untreated controls and pre-transplant samples is needed, CD8+ T cells from these DLBCL patients pre-blinatumomab contained very few naïve cells and were enriched for terminally differentiated effector cells. Conclusion: This pilot study shows that blinatumomab consolidation post auto-HCT is safe and well tolerated. MRD response to blinatumomab in all patients with MRD positive disease post auto-HCT is encouraging. Strategies to increase CD8/CD4 ratio and more cycles of consolidation in a larger randomized trial are needed to confirm the efficacy of consolidation with blinatumomab post auto-HCT. Finally, the unusually "high activation" immunophenotype (Teff/GB+) seen in CD8 T cells of DLBCL patients after auto-HCT (compared to those seen in resting peripheral blood) may both impact the response to blinatumomab and provide key insights into optimal timing for administration after auto-HCT. Disclosures Ghobadi: WuGen: Consultancy; Bristol Myers Squibb: Consultancy; Kite: Consultancy, Research Funding; EUSA: Consultancy; Amgen: Consultancy, Research Funding. Mehta-Shah:Bristol Myers-Squibb: Research Funding; Karyopharm Therapeutics: Consultancy; Corvus: Research Funding; Genetech/Roche: Research Funding; Innate Pharmaceuticals: Research Funding; Kyowa Hakko Kirin: Consultancy; Celgene: Research Funding; Verastem: Research Funding; C4 Therapeutics: Consultancy. Kahl:Celgene Corporation: Consultancy; AstraZeneca Pharmaceuticals LP: Consultancy, Membership on an entity's Board of Directors or advisory committees; Pharmacyclics LLC: Consultancy; Roche Laboratories Inc: Consultancy; BeiGene: Consultancy, Membership on an entity's Board of Directors or advisory committees, Research Funding; Janssen: Consultancy, Membership on an entity's Board of Directors or advisory committees; Acerta: Consultancy, Research Funding; ADC Therapeutics: Consultancy, Membership on an entity's Board of Directors or advisory committees, Research Funding; Genentech: Consultancy; AbbVie: Consultancy. DiPersio:Magenta Therapeutics: Membership on an entity's Board of Directors or advisory committees.

A semi-automatic cell type annotation method for single-cell RNA sequencing dataset.

Single-cell RNA sequencing (scRNA-seq) has been widely applied to provide insights into the cell-by-cell expression difference in a given bulk sample. Accordingly, numerous analysis methods have been developed. As it involves simultaneous analyses of many cell and genes, efficiency of the methods is crucial. The conventional cell type annotation method is laborious and subjective. Here we propose a semi-automatic method that calculates a normalized score for each cell type based on user-supplied cell type–specific marker gene list. The method was applied to a publicly available scRNA-seq data of mouse cardiac non-myocyte cell pool. Annotating the 35 t-stochastic neighbor embedding clusters into 12 cell types was straightforward, and its accuracy was evaluated by constructing co-expression network for each cell type. Gene Ontology analysis was congruent with the annotated cell type and the corollary regulatory network analysis showed upstream transcription factors that have well supported literature evidences. The source code is available as an R script upon request.

Identification of the key gene and pathways associated with osteoarthritis via single-cell RNA sequencing on synovial fibroblasts.

Osteoarthritis (OA) is a chronic degenerative joint disease with its onset closely related to the growth of synovial fibroblasts (SFs), yet the genes involved in are few reported. In our study, we aimed to identify the OA-associated key gene and pathways via the single-cell RNA sequencing (scRNA-seq) analysis on SFs.scRNA-seq data of SFs from OA sufferers were accessed from GEO database, then the genes involved in were subjected to principal component analysis (PCA) and T-Stochastic Neighbor Embedding (TSNE) Analysis. GO and KEGG enrichment analyses were performed to find the most enriched functions and pathways associated with marker genes and a PPI network was constructed to identify the key gene associated with OA occurrence.Findings revealed that marker genes in three cell types identified by TSNE were mainly activated in pathways firmly related to fibroblasts growth, such as extracellular matrix, immune and cell adhesion molecule binding-associated functions and pathways. Moreover, fibronectin1 (FN1) was validated as the key gene that was tightly related to the growth of SFs, as well as had the potential to play a key role in OA occurrence.Our study explored the key gene and pathways associated with OA occurrence, which were of great value in further investigation of OA diagnosis as well as pathogenesis.

Fast t-SNE algorithm with forest of balanced LSH trees and hybrid computation of repulsive forces

An acceleration of the well-known t-Stochastic Neighbor Embedding (t-SNE) (Hinton and Roweis, 2003; Maaten and Hinton, 2008) algorithm, probably the best (nonlinear) dimensionality reduction and visualization method, is proposed in this article.By using a specially-tuned forest of balanced trees constructed via locality sensitive hashing is improved significantly upon the results presented in Maaten (2014), achieving a complexity significantly closer to true O(nlogn), and vastly improving behavior for huge numbers of instances and attributes. Such acceleration removes the necessity to use PCA to reduce dimensionality before the start of t-SNE.Additionally, a fast hybrid method for repulsive forces computation (a part of the t-SNE algorithm), which is currently the fastest method known, is proposed.A parallelized version of our algorithm, characterized by a very good speedup factor, is proposed.

Condition Monitoring of Chain Sprocket Drive System Based on IoT Device and Convolutional Neural Network

This paper proposes a condition monitoring method for the early defect detection in a chain sprocket drive (CSD) system and classification of fault types before a catastrophic failure occurs. In the operation of a CSD system, early defect detection is very useful in preventing system failure. In this work, eight fault types associated with the CSD system components, such as the gear tooth, bearings, and drive motor shaft, were arbitrarily damaged and incorporated into the CSD system. To detect the fault signals during the CSD system operation, the vibration was measured using an Internet of Things (IoT) device, which features a wireless MEMS accelerometer, Bluetooth function, Wi-Fi function, and battery. The IoT device was mounted on the gearbox housing. The measured one-dimensional vibration time-series was transformed into time-scale images using continuous wavelet transform (CWT). A convolution neural network (CNN) was employed to extract deep features embedded in the images, which are closely related to fault types. To update the learning parameters of the CNN, the RMSprop learning algorithm was applied, and the CNN was trained using 500 image samples. Multiple-classification performance of the trained network was tested using 100 image samples. Feature maps for different fault types were obtained from the final CNN convolution layer. For the visualization of fault types, t-stochastic neighbor embedding was employed and applied to the feature maps to convert high-dimensional data into two-dimensional data. Two-dimensional features enabled excellent classification of the eight fault types and one normal type.

A comparative user study of visualization techniques for cluster analysis of multidimensional data sets

This article presents an empirical user study that compares eight multidimensional projection techniques for supporting the estimation of the number of clusters, [Formula: see text], embedded in six multidimensional data sets. The selection of the techniques was based on their intended design, or use, for visually encoding data structures, that is, neighborhood relations between data points or groups of data points in a data set. Concretely, we study: the difference between the estimates of [Formula: see text] as given by participants when using different multidimensional projections; the accuracy of user estimations with respect to the number of labels in the data sets; the perceived usability of each multidimensional projection; whether user estimates disagree with [Formula: see text] values given by a set of cluster quality measures; and whether there is a difference between experienced and novice users in terms of estimates and perceived usability. The results show that: dendrograms (from Ward’s hierarchical clustering) are likely to lead to estimates of [Formula: see text] that are different from those given with other multidimensional projections, while Star Coordinates and Radial Visualizations are likely to lead to similar estimates; t-Stochastic Neighbor Embedding is likely to lead to estimates which are closer to the number of labels in a data set; cluster quality measures are likely to produce estimates which are different from those given by users using Ward and t-Stochastic Neighbor Embedding; U-Matrices and reachability plots will likely have a low perceived usability; and there is no statistically significant difference between the answers of experienced and novice users. Moreover, as data dimensionality increases, cluster quality measures are likely to produce estimates which are different from those perceived by users using any of the assessed multidimensional projections. It is also apparent that the inherent complexity of a data set, as well as the capability of each visual technique to disclose such complexity, has an influence on the perceived usability.

Deep Double Supervised Embedding Neural Network Enhancing Class Separation for Visual High-Dimensional Industrial Process Monitoring

Visual process monitoring is the application of a visualization method to map the real-time operating information of an industrial process to a 2-D map, followed by process monitoring. However, owing to the complexity of industrial production processes and the complex correlations among industrial process variables, the structure and distribution of high-dimensional industrial data are very complicated. Therefore, a general visualization method cannot effectively separate the different fault data in a 2-D map for process monitoring. Accordingly, in this article, a deep double supervised embedding neural network (DDSE) is proposed for visualizing high-dimensional industrial data. The DDSE consists of two supervised deep neural networks: a deep class centres uniform distribution neural network (DCCUD), and a deep supervised t-stochastic neighbor embedding neural network (DSSNE). The DCCUD maps the high-dimensional industrial data to a new feature space in which the class centres obey a uniform distribution, promoting a good and separable situation for subsequent visualization procedures. The DSSNE then maps these high-dimensional features into a 2-D space. The training of the DDSE can be conducted through pre-training and fine-tuning. A proposed visual process monitoring approach combines the DDSE with the local outlier factor and k-nearest neighbor approaches. The proposed approach is tested on a Tennessee Eastman process, and the results illustrate that the proposed approach outperforms traditional methods in terms of visualization and visual process monitoring.

Radiomics in Cardiac MRI: Sisyphean Struggle or Close to the Summit of Olympus?

Radiomics in Cardiac MRI: Sisyphean Struggle or Close to the Summit of Olympus?