SVD-AE: An asymmetric autoencoder with SVD regularization for multivariate time series anomaly detection.

Anomaly detection in multivariate time series is of critical importance in many real-world applications, such as system maintenance and Internet monitoring. In this article, we propose a novel unsupervised framework called SVD-AE to conduct anomaly detection in multivariate time series. The core idea is to fuse the strengths of both SVD and autoencoder to fully capture complex normal patterns in multivariate time series. An asymmetric autoencoder architecture is proposed, where two encoders are used to capture features in time and variable dimensions and a shared decoder is used to generate reconstructions based on latent representations from both dimensions. A new regularization based on singular value decomposition theory is designed to force each encoder to learn features in the corresponding axis with mathematical supports delivered. A specific loss component is further proposed to align Fourier coefficients of inputs and reconstructions. It can preserve details of original inputs, leading to enhanced feature learning capability of the model. Extensive experiments on three real world datasets demonstrate the proposed algorithm can achieve better performance on multivariate time series anomaly detection tasks under highly unbalanced scenarios compared with baseline algorithms.

Variation in structural and in vitro starch digestion of pulse cotyledon cells imposed by temperature-pressure-moisture combinations.

Starch digestibility in whole pulses is affected by food structural characteristics, which in turn can be modulated by processing methods. In present study, high-pressure steam (HPS) and hydrothermal treatment (HT) with different moisture content were applied to clarify the mechanisms of processing variables affecting in vitro starch digestibility in pulse cells. Based on thermal and X-ray results, the relative crystallinity of cells decreased after HPS and HT treatments. However, HPS-treated cells under higher (>50%) moisture content showed insignificant discrepancies in crystallinity than HT samples. Starch digestion in HPS-treated cells increased with higher moisture content but was still lower than in HT samples. Results of FITC-dextran diffusion and methyl esterification of cell walls indicated that cells with higher wall permeability exhibited relatively higher starch digestibility. This study suggests that the enzyme susceptibility to starch in cells is dominantly influenced by cell wall structure, which could be optimized through processing variables.

Development of a chemiluminescent aptasensor for ultrasensitive and selective detection of aflatoxin B1 in peanut and milk.

More and more attention about food safety leads to a research hotspot to develop new detection methods for food contaminant. To address the problems of serious interference and low sensitivity, a chemiluminescent aptasensor for the detection of aflatoxin B1(AFB1) in food was developed in this paper. It is based on horseradish peroxidase (HRP) catalyze the luminol chemiluminescence reaction. The hybridization chain reaction (HCR) signal amplification strategy has been used to improve the detection sensitivity. Magnetic separation could further reduce background signal obviously at the same time. AFB1 as a model of analyte to test the capability of our developed assay system. Under the optimal experimental conditions, CL intensity showed a good linear correlation with the concentrations of AFB1 ranging from 0.5 to 40 ng mL-1. The limit of detection was estimated 0.2 ng mL-1 based on 3 times of the signal-to-noise ratio which is lower than those of the previously reported sensors. It could be used to detect AFB1 content in real samples, such as peanuts and milk which were purchased in local supermarket. The results proved that the sensing system has good anti-interference and selectivity. In all, it has potential for practical application in food safety field.

A highly sensitive and low-background fluorescence assay for pesticides residues based on hybridization chain reaction amplification assisted by magnetic separation.

Due to the concern over food safety, it is important to detect the pesticides residues in agricultural products. Here, a highly sensitive and low background fluorescent strategy for the detection of pesticides residues has been developed. The fluorescence intensity of N-methyl mesoporphyrin IX (NMM) binding G-quadruplex could be turn off because of inhibiting effect of the pesticides on the acetylcholinesterase (AChE) activity. For that, four single-stranded DNAs (named linker, trigger, H1 and H2, respectively) are rational designed and T-Hg-T mismatches duplex DNAs as a recognizer combined with the separation of magnetic beads. The design of hybridization chain reaction (HCR) amplification strategy assisted by magnetic separation has been adopted to improve the detection sensitivity. In the presence of pesticides, the amount of the thiol group generated by hydrolysis reaction of acetylcholine (ACh) is reduced, lead to release of less trigger DNA. Therefor subsequent HCR process is retarded with decreased fluorescence intensity. The reduced fluorescence intensity has a quantitative relationship with the pesticide concentration. The limit of detection of chlorpyrifos was estimated to be 2.0 ng ml-1. It has been applied to detect the pesticides residues in real samples.

A label-free, versatile and low-background chemiluminescence aptasensing strategy based on gold nanocluster catalysis combined with the separation of magnetic beads.

A label-free, versatile and low-background chemiluminescence (CL) sensing strategy based on gold nanocluster catalysis combined with the separation of magnetic beads (MBs) was developed. Kanamycin was selected as the target analyte to exhibit the analytical performance of this platform. Two single-stranded DNA (named DNA1 and DNA2) are ingeniously designed. DNA1, containing an aptamer sequence of the targets, was firstly immobilized on the MBs which were modified with many amino groups by amidation reaction. DNA2 consists of 30 repeat adenosine bases (A30) at the 5' terminal which were used to prepare AuNCs by a UV-light-assisted method and a 12 nucleotide sequence at the 3' terminal which can easily hybridize with DNA1 to form a partly complementary double-stranded structure. In the presence of targets, the aptamer modified on MBs would combine with targets and lead to release the prepared DNA-templated AuNCs. After the magnetic separation, enrichment AuNCs in the supernatant can catalyze the CL substrate to produce a strong CL signal. The well-designed CL sensing strategy exhibited a low detection limit of 0.035 nM for kanamycin, and it also showed good selectivity and stability. Most importantly, different targets can be analyzed only by changing the aptamer sequence that is immobilized on the MBs. Therefore, the strategy we proposed here has provided a versatile sensing platform for sensitively detecting various biomolecules at low levels.