A Comparative Evaluation of Unsupervised Anomaly Detection Algorithms for Multivariate Data

Anomaly detection is the process of identifying unexpected items or events in datasets, which differ from the norm. In contrast to standard classification tasks, anomaly detection is often applied on unlabeled data, taking only the internal structure of the dataset into account. This challenge is known as unsupervised anomaly detection and is addressed in many practical applications, for example in network intrusion detection, fraud detection as well as in the life science and medical domain. Dozens of algorithms have been proposed in this area, but unfortunately the research community still lacks a comparative universal evaluation as well as common publicly available datasets. These shortcomings are addressed in this study, where 19 different unsupervised anomaly detection algorithms are evaluated on 10 different datasets from multiple application domains. By publishing the source code and the datasets, this paper aims to be a new well-funded basis for unsupervised anomaly detection research. Additionally, this evaluation reveals the strengths and weaknesses of the different approaches for the first time. Besides the anomaly detection performance, computational effort, the impact of parameter settings as well as the global/local anomaly detection behavior is outlined. As a conclusion, we give an advise on algorithm selection for typical real-world tasks.     

Competition for Mitogens Regulates Spermatogenic Stem Cell Homeostasis in an Open Niche

1. Download high-res image (317KB)
2. Download full-size image

     

Crystallographic and spectroscopic studies of peroxide-derived myoglobin compound II and occurrence of protonated FeIV O

High resolution crystal structures of myoglobin in the pH range 5.2-8.7 have been used as models for the peroxide-derived compound II intermediates in heme peroxidases and oxygenases. The observed Fe-O bond length (1.86-1.90 A) is consistent with that of a single bond. The compound II state of myoglobin in crystals was controlled by single-crystal microspectrophotometry before and after synchrotron data collection. We observe some radiation-induced changes in both compound II (resulting in intermediate H) and in the resting ferric state of myoglobin. These radiation-induced states are quite unstable, and compound II and ferric myoglobin are immediately regenerated through a short heating above the glass transition temperature (<1 s) of the crystals. It is unclear how this influences our compound II structures compared with the unaffected compound II, but some crystallographic data suggest that the influence on the Fe-O bond distance is minimal. Based on our crystallographic and spectroscopic data we suggest that for myoglobin the compound II intermediate consists of an Fe(IV)-O species with a single bond. The presence of Fe(IV) is indicated by a small isomer shift of delta = 0.07 mm/s from M?ssbauer spectroscopy. Earlier quantum refinements (crystallographic refinement where the molecular-mechanics potential is replaced by a quantum chemical calculation) and density functional theory calculations suggest that this intermediate H species is protonated.     

Involvement of an autocrine stromal cell derived factor-1/CXCR4 system on the distant metastasis of human oral squamous cell carcinoma

We have previously shown that a stromal cell-derived factor-1 (SDF-1; CXCL12)/CXCR4 system is involved in the establishment of lymph node metastasis, but not in that of distant metastasis, in oral squamous cell carcinoma (SCC). In this study, we investigated the role of the autocrine SDF-1/CXCR4 system, with a focus on distant metastasis in oral SCC cells. The immunohistochemical staining of SDF-1 and CXCR4 using primary oral SCCs and metastatic lymph nodes showed a significantly higher number of SDF-1-positive cases among the metastatic lymph nodes than among the primary oral SCCs, which was associated with a poor survival rate among those of the former group. The forced expression of SDF-1 in B88 cells, which exhibit functional CXCR4 and lymph node metastatic potential (i.e., the autocrine SDF-1/CXCR4 system), conferred enhanced cell motility and anchorage-independent growth potential onto the cells. Orthotopic inoculation of the transfectant into nude mice was associated with an increase in the number of metastatic lymph nodes and more aggressive metastatic foci in the lymph nodes. Furthermore, the SDF-1 transfectant (i.e., the autocrine SDF-1/CXCR4 system) exhibited dramatic metastasis to the lung after i.v. inoculation, whereas the mock transfectant (i.e., the paracrine SDF-1/CXCR4 system) did not. Under the present conditions, AMD3100, a CXCR4 antagonist, significantly inhibited the lung metastasis of the SDF-1 transfectant, ameliorated body weight loss, and improved the survival rate of tumor-bearing nude mice. These results suggested that, in cases of oral SCC, the paracrine SDF-1/CXCR4 system potentiates lymph node metastasis, but distant metastasis might require the autocrine SDF-1/CXCR4 system.     

Degradation of nuclear DNA by DNase II-like acid DNase in cortical fiber cells of mouse eye lens

The eye lens is composed of fiber cells that differentiate from epithelial cells on its anterior surface. In concert with this differentiation, a set of proteins essential for lens function is synthesized, and the cellular organelles are degraded. DNase II-like acid DNase, also called DNase IIbeta, is specifically expressed in the lens, and degrades the DNA in the lens fiber cells. Here we report that DNase II-like acid DNase is synthesized as a precursor with a signal sequence, and is localized to lysosomes. DNase II-like acid DNase mRNA was found in cortical fiber cells but not epithelial cells, indicating that its expression is induced during the differentiation of epithelial cells into fiber cells. Immunohistochemical and immunocytochemical analyses indicated that DNase II-like acid DNase was colocalized with Lamp-1 in the lysosomes of fiber cells in a relatively narrow region bordering the organelle-free zone, and was often found in degenerating nuclei. A comparison by microarray analysis of the gene expression profiles between epithelial and cortical fiber cells of young mouse lens indicated that some genes for lysosomal enzymes (cathepsins and lipases) were strongly expressed in the fiber cells. These results suggest that the lysosomal system plays a role in the degradation of cellular organelles during lens cell differentiation.     

Changes in sleep quality of athletes under normobaric hypoxia equivalent to 2,000-m altitude: a polysomnographic study.

This study evaluated the sleep quality of athletes in normobaric hypoxia at a simulated altitude of 2,000 m. Eight male athletes slept in normoxic condition (NC) and hypoxic conditions equivalent to those at 2,000-m altitude (HC). Polysomnographic recordings of sleep included the electroencephalogram (EEG), electrooculogram, chin surface electromyogram, and electrocardiogram. Thoracic and abdominal motion, nasal and oral airflow, and arterial blood oxygen saturation (Sa(O(2))) were also recorded. Standard visual sleep stage scoring and fast Fourier transformation analyses of the EEG were performed on 30-s epochs. Subjective sleepiness and urinary catecholamines were also monitored. Mean Sa(O(2)) decreased and respiratory disturbances increased with HC. The increase in respiratory disturbances was significant, but the increase was small and subclinical. The duration of slow-wave sleep (stage 3 and 4) and total delta power (<3 Hz) of the all-night non-rapid eye movement sleep EEG decreased for HC compared with NC. Subjective sleepiness and amounts of urinary catecholamines did not differ between the conditions. These results indicate that acute exposure to normobaric hypoxia equivalent to that at 2,000-m altitude decreased slow-wave sleep in athletes, but it did not change subjective sleepiness or amounts of urinary catecholamines.     

Measurement of fasting serum apoB-48 levels in normolipidemic and hyperlipidemic subjects by ELISA

The present study was designed to evaluate the metabolism of chylomicron and chylomicron remnants by measuring serum apolipoprotein B-48 (apoB-48) levels in 335 normolipidemic and 253 hyperlipidemic subjects using a novel ELISA system. The distribution of fasting serum apoB-48 levels in normolipidemic subjects varied widely, ranging from <1 to >24 microgram/ml (mean, 5.2 +/- 3.8 microgram/ml; median, 3.9 microgram/ml). Serum apoB-48 levels correlated with serum triglyceride (TG) concentrations (r = 0.45, P < 0.001), but not with total cholesterol levels. Serum apoB-48 levels were 7 to 18 times higher in patients with Type I, Type V, and Type III hyperlipidemia, and only slightly higher in patients with Type IIa, Type IIb, and Type IV hyperlipidemia, compared with normolipidemic subjects. The calculated apoB-48/TG ratio was elevated only in patients with dysbetalipoproteinemia (apoE2/2 phenotype). In normolipidemic subjects, oral fat loading resulted in about 2-fold increase in serum apoB-48 levels, with a peak level recorded at 3-4 h postloading, and then returned to the baseline level within 6 h. On the other hand, in patients with dysbetalipoproteinemia, serum apoB-48 levels did not change considerably. Our results indicate that serum apoB-48 is a very useful parameter for evaluating lipoprotein metabolism in exogenous pathways.     

Contribution of micro-organisms to the carbon dynamics in black spruce (Picea mariana) forest soil in Canada

In order to clarify the role of micro-organisms in the carbon cycle of the boreal forest ecosystem, the vertical distribution of soil carbon, soil microbial biomass and respiratory activity was studied in a black spruce forest near Candle Lake in Saskatchewan, Canada. The total amount of carbon contained in moss and soil layers (to the depth of 50cm beneath the mineral soil surface) was 7.2kgm^–2, about 47% of which was in the L and FH horizons of the soil. Soil microbial biomass per dry weight of soil was largest in the L horizon, while the biomass per ground area was largest in the FH horizon. Soil respiration rate, measured using a portable infrared gas analyzer, was highest in the FH horizon, exceeding 50% of the total soil respiration. Low but significant CO₂ emission was detected even in deeper soil horizon (E horizon). We also examined the respiration rate of cut roots and the effect of root excision on respiration. The contribution of root respiration to total soil respiration, calculated from root biomass and respiration rate of cut roots, was about 54%. The amount of carbon evolved through microbial respiration during the snow-free season (June–October) was estimated as 221gCm^–2. Micro-organisms in the L horizon showed high respiratory activity as compared with those in deeper soil horizons.     

Rapid and effective detection of anthrax spores in soil by PCR

AIMS: To detect Bacillus anthracis DNA from soil using rapid and simple procedures. METHODS AND RESULTS: Various amounts of B. anthracis Pasteur II spores were added artificially to 1 g of soil, which was then washed with ethanol and sterile water. Enrichment of the samples in trypticase soy broth was performed twice. A DNA template was prepared from the second enrichment culture using a FastPrep instrument. The template was then used for nested and real-time polymerase chain reaction (PCR) with B. anthracis-specific primers, to confirm the presence of B. anthracis chromosomal DNA and the pXO1/pXO2 plasmids. CONCLUSIONS: One cell of B. anthracis in 1 g of soil could be detected by nested and real-time PCR. The usefulness of the PCR method using field samples was also confirmed. SIGNIFICANCE AND IMPACT OF THE STUDY: The results indicate that this could be a useful method for detecting anthrax-spore contaminated soil with high sensitivity. Its application could have great impact on the progress of epidemiological surveillance.     

Purification and characterization of an aldehyde oxidase from Pseudomonas sp. KY 4690

An aldehyde oxidase, which oxidizes various aliphatic and aromatic aldehydes using O(2) as an electron acceptor, was purified from the cell-free extracts of Pseudomonas sp. KY 4690, a soil isolate, to an electrophoretically homogeneous state. The purified enzyme had a molecular mass of 132 kDa and consisted of three non-identical subunits with molecular masses of 88, 39, and 18 kDa. The absorption spectrum of the purified enzyme showed characteristics of an enzyme belonging to the xanthine oxidase family. The enzyme contained 0.89 mol of flavin adenine dinucleotide, 1.0 mol of molybdenum, 3.6 mol of acid-labile sulfur, and 0.90 mol of 5'-CMP per mol of enzyme protein, on the basis of its molecular mass of 145 kDa. Molecular oxygen served as the sole electron acceptor. These results suggest that aldehyde oxidase from Pseudomonas sp. KY 4690 is a new member of the xanthine oxidase family and might contain 1 mol of molybdenum-molybdpterin-cytosine dinucleotide, 1 mol of flavin adenine dinucleotide, and 2 mol of [2Fe-2S] clusters per mol of enzyme protein. The enzyme showed high reaction rates toward various aliphatic and aromatic aldehydes and high thermostability.