Anomaly Detection Using an Ensemble of Multi-Point LSTMs.

As technologies for storing time-series data such as smartwatches and smart factories become common, we are collectively accumulating a great deal of time-series data. With the accumulation of time-series data, the importance of time-series abnormality detection technology that detects abnormal patterns such as Cyber-Intrusion Detection, Fraud Detection, Social Networks Anomaly Detection, and Industrial Anomaly Detection is emerging. In the past, time-series anomaly detection algorithms have mainly focused on processing univariate data. However, with the development of technology, time-series data has become complicated, and corresponding deep learning-based time-series anomaly detection technology has been actively developed. Currently, most industries rely on deep learning algorithms to detect time-series anomalies. In this paper, we propose an anomaly detection algorithm with an ensemble of multi-point LSTMs that can be used in three cases of time-series domains. We propose our anomaly detection model that uses three steps. The first step is a model selection step, in which a model is learned within a user-specified range, and among them, models that are most suitable are automatically selected. In the next step, a collected output vector from M LSTMs is completed by stacking ensemble techniques of the previously selected models. In the final step, anomalies are finally detected using the output vector of the second step. We conducted experiments comparing the performance of the proposed model with other state-of-the-art time-series detection deep learning models using three real-world datasets. Our method shows excellent accuracy, efficient execution time, and a good F1 score for the three datasets, though training the LSTM ensemble naturally requires more time.

Feature selection using distributions of orthogonal PLS regression vectors in spectral data.

Feature selection, which is important for successful analysis of chemometric data, aims to produce parsimonious and predictive models. Partial least squares (PLS) regression is one of the main methods in chemometrics for analyzing multivariate data with input X and response Y by modeling the covariance structure in the X and Y spaces. Recently, orthogonal projections to latent structures (OPLS) has been widely used in processing multivariate data because OPLS improves the interpretability of PLS models by removing systematic variation in the X space not correlated to Y. The purpose of this paper is to present a feature selection method of multivariate data through orthogonal PLS regression (OPLSR), which combines orthogonal signal correction with PLS. The presented method generates empirical distributions of features effects upon Y in OPLSR vectors via permutation tests and examines the significance of the effects of the input features on Y. We show the performance of the proposed method using a simulation study in which a three-layer network structure exists in compared with the false discovery rate method. To demonstrate this method, we apply it to both real-life NIR spectra data and mass spectrometry data.

A biplot correlation range for group-wise metabolite selection in mass spectrometry.

BACKGROUND: Analytic methods are available to acquire extensive metabolic information in a cost-effective manner for personalized medicine, yet disease risk and diagnosis mostly rely upon individual biomarkers based on statistical principles of false discovery rate and correlation. Due to functional redundancies and multiple layers of regulation in complex biologic systems, individual biomarkers, while useful, are inherently limited in disease characterization. Data reduction and discriminant analysis tools such as principal component analysis (PCA), partial least squares (PLS), or orthogonal PLS (O-PLS) provide approaches to separate the metabolic phenotypes, but do not offer a statistical basis for selection of group-wise metabolites as contributors to metabolic phenotypes. METHODS: We present a dimensionality-reduction based approach termed 'biplot correlation range (BCR)' that uses biplot correlation analysis with direct orthogonal signal correction and PLS to provide the group-wise selection of metabolic markers contributing to metabolic phenotypes. RESULTS: Using a simulated multiple-layer system that often arises in complex biologic systems, we show the feasibility and superiority of the proposed approach in comparison of existing approaches based on false discovery rate and correlation. To demonstrate the proposed method in a real-life dataset, we used LC-MS based metabolomics to determine spectrum of metabolites present in liver mitochondria from wild-type (WT) mice and thioredoxin-2 transgenic (TG) mice. We select discriminatory variables in terms of increased score in the direction of class identity using BCR. The results show that BCR provides means to identify metabolites contributing to class separation in a manner that a statistical method by false discovery rate or statistical total correlation spectroscopy can hardly find in complex data analysis for predictive health and personalized medicine.

Characterizing exons and introns by regularity of nucleotide strings.

BACKGROUND: Translation of nucleotides into a numeric form has been approached in many ways and has allowed researchers to investigate the properties of protein-coding sequences and noncoding sequences. Typically, more pronounced long-range correlations and increased regularity were found in intron-containing genes and in non-transcribed regulatory DNA sequences, compared to cDNA sequences or intron-less genes. The regularity is assessed by spectral tools defined on numerical translates. In most popular approaches of numerical translation the resulting spectra depend on the assignment of numerical values to nucleotides. Our contribution is to propose and illustrate a spectra which remains invariant to the translation rules used in traditional approaches. RESULTS: We outline a methodology for representing sequences of DNA nucleotides as numeric matrices in order to analytically investigate important structural characteristics of DNA. This representation allows us to compute the 2-dimensional wavelet transformation and assess regularity characteristics of the sequence via the slope of the wavelet spectra. In addition to computing a global slope measure for a sequence, we can apply our methodology for overlapping sections of nucleotides to obtain an "evolutionary slope." To illustrate our methodology, we analyzed 376 gene sequences from the first chromosome of the honeybee. CONCLUSION: For the genes analyzed, we find that introns are significantly more regular (lead to more negative spectral slopes) than exons, which agrees with the results from the literature where regularity is measured on "DNA walks". However, unlike DNA walks where the nucleotides are assigned numerical values depending on nucleotide characteristics (purine-pyrimidine, weak-strong hydrogen bonds, keto-amino, etc.) or other spatial assignments, the proposed spectral tool is invariant to the assignment of nucleotides. Thus, ambiguity in numerical translation of nucleotides is eliminated.

Optimal ligand descriptor for pocket recognition based on the Beta-shape.

Structure-based virtual screening is one of the most important and common computational methods for the identification of predicted hit at the beginning of drug discovery. Pocket recognition and definition is frequently a prerequisite of structure-based virtual screening, reducing the search space of the predicted protein-ligand complex. In this paper, we present an optimal ligand shape descriptor for a pocket recognition algorithm based on the beta-shape, which is a derivative structure of the Voronoi diagram of atoms. We investigate six candidates for a shape descriptor for a ligand using statistical analysis: the minimum enclosing sphere, three measures from the principal component analysis of atoms, the van der Waals volume, and the beta-shape volume. Among them, the van der Waals volume of a ligand is the optimal shape descriptor for pocket recognition and best tunes the pocket recognition algorithm based on the beta-shape for efficient virtual screening. The performance of the proposed algorithm is verified by a benchmark test.

Serum metabolomics of slow vs. rapid motor progression Parkinson's disease: a pilot study.

Progression of Parkinson's disease (PD) is highly variable, indicating that differences between slow and rapid progression forms could provide valuable information for improved early detection and management. Unfortunately, this represents a complex problem due to the heterogeneous nature of humans in regards to demographic characteristics, genetics, diet, environmental exposures and health behaviors. In this pilot study, we employed high resolution mass spectrometry-based metabolic profiling to investigate the metabolic signatures of slow versus rapidly progressing PD present in human serum. Archival serum samples from PD patients obtained within 3 years of disease onset were analyzed via dual chromatography-high resolution mass spectrometry, with data extraction by xMSanalyzer and used to predict rapid or slow motor progression of these patients during follow-up. Statistical analyses, such as false discovery rate analysis and partial least squares discriminant analysis, yielded a list of statistically significant metabolic features and further investigation revealed potential biomarkers. In particular, N8-acetyl spermidine was found to be significantly elevated in the rapid progressors compared to both control subjects and slow progressors. Our exploratory data indicate that a fast motor progression disease phenotype can be distinguished early in disease using high resolution mass spectrometry-based metabolic profiling and that altered polyamine metabolism may be a predictive marker of rapidly progressing PD.

Metabolome-wide association study of neovascular age-related macular degeneration.

PURPOSE: To determine if plasma metabolic profiles can detect differences between patients with neovascular age-related macular degeneration (NVAMD) and similarly-aged controls. METHODS: Metabolomic analysis using liquid chromatography with Fourier-transform mass spectrometry (LC-FTMS) was performed on plasma samples from 26 NVAMD patients and 19 controls. Data were collected from mass/charge ratio (m/z) 85 to 850 on a Thermo LTQ-FT mass spectrometer, and metabolic features were extracted using an adaptive processing software package. Both non-transformed and log2 transformed data were corrected using Benjamini and Hochberg False Discovery Rate (FDR) to account for multiple testing. Orthogonal Partial Least Squares-Discriminant Analysis was performed to determine metabolic features that distinguished NVAMD patients from controls. Individual m/z features were matched to the Kyoto Encyclopedia of Genes and Genomes database and the Metlin metabolomics database, and metabolic pathways associated with NVAMD were identified using MetScape. RESULTS: Of the 1680 total m/z features detected by LC-FTMS, 94 unique m/z features were significantly different between NVAMD patients and controls using FDR (q = 0.05). A comparison of these features to those found with log2 transformed data (n = 132, q = 0.2) revealed 40 features in common, reaffirming the involvement of certain metabolites. Such metabolites included di- and tripeptides, covalently modified amino acids, bile acids, and vitamin D-related metabolites. Correlation analysis revealed associations among certain significant features, and pathway analysis demonstrated broader changes in tyrosine metabolism, sulfur amino acid metabolism, and amino acids related to urea metabolism. CONCLUSIONS: These data suggest that metabolomic analysis can identify a panel of individual metabolites that differ between NVAMD cases and controls. Pathway analysis can assess the involvement of certain metabolic pathways, such as tyrosine and urea metabolism, and can provide further insight into the pathophysiology of AMD.

Multifractal analysis for nutritional assessment.

The concept of multifractality is currently used to describe self-similar and complex scaling properties observed in numerous biological signals. Fractals are geometric objects or dynamic variations which exhibit some degree of similarity (irregularity) to the original object in a wide range of scales. This approach determines irregularity of biologic signal as an indicator of adaptability, the capability to respond to unpredictable stress, and health. In the present work, we propose the application of multifractal analysis of wavelet-transformed proton nuclear magnetic resonance ((1)H NMR) spectra of plasma to determine nutritional insufficiency. For validation of this method on (1)H NMR signal of human plasma, standard deviation from classical statistical approach and Hurst exponent (H), left slope and partition function from multifractal analysis were extracted from (1)H NMR spectra to test whether multifractal indices could discriminate healthy subjects from unhealthy, intensive care unit patients. After validation, the multifractal approach was applied to spectra of plasma from a modified crossover study of sulfur amino acid insufficiency and tested for associations with blood lipids. The results showed that standard deviation and H, but not left slope, were significantly different for sulfur amino acid sufficiency and insufficiency. Quadratic discriminant analysis of H, left slope and the partition function showed 78% overall classification accuracy according to sulfur amino acid status. Triglycerides and apolipoprotein C3 were significantly correlated with a multifractal model containing H, left slope, and standard deviation, and cholesterol and high-sensitivity C-reactive protein were significantly correlated to H. In conclusion, multifractal analysis of (1)H NMR spectra provides a new approach to characterize nutritional status.

High-performance metabolic profiling of plasma from seven mammalian species for simultaneous environmental chemical surveillance and bioeffect monitoring.

High-performance metabolic profiling (HPMP) by Fourier-transform mass spectrometry coupled to liquid chromatography gives relative quantification of thousands of chemicals in biologic samples but has had little development for use in toxicology research. In principle, the approach could be useful to detect complex metabolic response patterns to toxicologic exposures and to detect unusual abundances or patterns of potentially toxic chemicals. As an initial study to develop these possible uses, we applied HPMP and bioinformatics analysis to plasma of humans, rhesus macaques, marmosets, pigs, sheep, rats and mice to determine: (1) whether more chemicals are detected in humans living in a less controlled environment than captive species and (2) whether a subset of plasma chemicals with similar inter-species and intra-species variation could be identified for use in comparative toxicology. Results show that the number of chemicals detected was similar in humans (3221) and other species (range 2537-3373). Metabolite patterns were most similar within species and separated samples according to family and order. A total of 1485 chemicals were common to all species; 37% of these matched chemicals in human metabolomic databases and included chemicals in 137 out of 146 human metabolic pathways. Probability-based modularity clustering separated 644 chemicals, including many endogenous metabolites, with inter-species variation similar to intra-species variation. The remaining chemicals had greater inter-species variation and included environmental chemicals as well as GSH and methionine. Together, the data suggest that HPMP provides a platform that can be useful within human populations and controlled animal studies to simultaneously evaluate environmental exposures and biological responses to such exposures.

Metabolic effects of albumin therapy in acute lung injury measured by proton nuclear magnetic resonance spectroscopy of plasma: a pilot study.

OBJECTIVE: Improved means to monitor and guide interventions could be useful in the intensive care unit. Metabolomic analysis with bioinformatics is used to understand mechanisms and identify biomarkers of disease development and progression. This pilot study evaluated plasma proton nuclear magnetic resonance spectroscopy as a means to monitor metabolism following albumin administration in acute lung injury patients. DESIGN: This study was conducted on plasma samples from six albumin-treated and six saline-treated patients from a larger double-blind trial. The albumin group was administered 25 g of 25% human albumin in 0.9% saline every 8 hrs for a total of nine doses over 72 hrs. A 0.9% concentration of saline was used as a placebo. Blood samples were collected immediately before, 1 hr after, and 4 hrs after the albumin/saline administration for the first, fourth, and seventh doses (first dose of each day for 3 days). Samples were analyzed by proton nuclear magnetic resonance spectroscopy, and spectra were analyzed by principal component analysis and biostatistical methods. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: After 1 day of albumin therapy, changes in small molecules, including amino acids and plasma lipids, were evident with principal component analysis. Differences remained 3 days after the last albumin administration. Analysis of data along with spectra from healthy controls showed that spectra for patients receiving albumin had a trajectory toward the spectra observed for healthy individuals while those of the placebo controls did not. CONCLUSION: The data suggest that metabolic changes detected by proton nuclear magnetic resonance spectroscopy and the bioinformatics tool may be a useful approach to clinical research, especially in acute lung injury.

A sulfur amino acid-free meal increases plasma lipids in humans.

The content of sulfur amino acid (SAA) in a meal affects postprandial plasma cysteine concentrations and the redox potential of cysteine/cystine. Because such changes can affect enzyme, transporter, and receptor activities, meal content of SAA could have unrecognized effects on metabolism during the postprandial period. This pilot study used proton NMR ((1)H-NMR) spectroscopy of human plasma to test the hypothesis that dietary SAA content changes macronutrient metabolism. Healthy participants (18-36 y, 5 males and 3 females) were equilibrated for 3 d to adequate SAA, fed chemically defined meals without SAA for 5 d (depletion), and then fed isoenergetic, isonitrogenous meals containing 56 mg·kg(-1)·d(-1) SAA for 4.5 d (repletion). On the first and last day of consuming the chemically defined meals, a morning meal containing 60% of the daily food intake was given and plasma samples were collected over an 8-h postprandial time course for characterization of metabolic changes by (1)H-NMR spectroscopy. SAA-free food increased peak intensity in the plasma (1)H-NMR spectra in the postprandial period. Orthogonal signal correction/partial least squares-discriminant analysis showed changes in signals associated with lipids, some amino acids, and lactate, with notable increases in plasma lipid signals (TG, unsaturated lipid, cholesterol). Conventional lipid analyses confirmed higher plasma TG and showed an increase in plasma concentration of the lipoprotein lipase inhibitor, apoC-III. The results show that plasma (1)H-NMR spectra can provide useful macronutrient profiling following a meal challenge protocol and that a single meal with imbalanced SAA content alters postprandial lipid metabolism.

Breath analysis by mass spectrometry: a new tool for breast cancer detection?

Breath analysis has received attention as a noninvasive diagnostic tool with increasing research into its potential usefulness. We are investigating the utility of the analysis of breath volatile organic compounds (VOCs) as an effective modality for breast cancer (BC) detection and monitoring by collecting breath samples with a simple portable device to determine whether BC patients have breath VOCs distinct from those in healthy volunteers. We prospectively enrolled 20 healthy volunteers and 20 newly diagnosed stage II-IV BC patients. The study subjects deeply exhaled into a commercially available Teflon/valved breath sampler equipped with a rapid passive diffusive sampler five times at 5-minute intervals trapping alveolar breath VOCs. The exhaled breath samples were analyzed by thermal desorption/gas chromatography/mass spectrometry monitoring 383 VOCs in the breath of both populations. Our results indicate that aggregate low-dimensional summaries and compound quantities result in specific patterns that can confirm BC. We found a definite clustering of the presence of BC from cancer-free points. Overall sensitivity was 72 per cent and specificity was 64 per cent resulting in a correct classification rate of approximately 77 per cent. Our data show promising evidence that BC patients can be differentiated from healthy volunteers through distinct breath VOCs.

Orbital trauma.

Assessment of the orbit for orbital trauma is best achieved expeditiously with CT in the determination of extent of injury and the presence of foreign body. MR imaging has a limited role but is valuable in examining the optic nerve and globe for injury and has proven to be an adjunct modality in the assessment of orbital injury.