Deterministic convergence of chaos injection-based gradient method for training feedforward neural networks

It has been shown that, by adding a chaotic sequence to the weight update during the training of neural networks, the chaos injection-based gradient method (CIBGM) is superior to the standard backpropagation algorithm. This paper presents the theoretical convergence analysis of CIBGM for training feedforward neural networks. We consider both the case of batch learning as well as the case of online learning. Under mild conditions, we prove the weak convergence, i.e., the training error tends to a constant and the gradient of the error function tends to zero. Moreover, the strong convergence of CIBGM is also obtained with the help of an extra condition. The theoretical results are substantiated by a simulation example.     

Analysis of a Panel of 48 Cytokines in BAL Fluids Specifically Identifies IL-8 Levels as the Only Cytokine that Distinguishes Controlled Asthma from Uncontrolled Asthma,and Correlates Inversely with FEV1

We sought to identify cells and cytokines in bronchoalveolar lavage (BAL) fluids that distinguish asthma from healthy control subjects and those that distinguish controlled asthma from uncontrolled asthma. Following informed consent, 36 human subjects were recruited for this study. These included 11 healthy control subjects, 15 subjects with controlled asthma with FEV₁≥80% predicted and 10 subjects with uncontrolled asthma with FEV₁ <80% predicted. BAL fluid was obtained from all subjects. The numbers of different cell types and the levels of 48 cytokines were measured in these fluids. Compared to healthy control subjects, patients with asthma had significantly more percentages of eosinophils and neutrophils, IL-1RA, IL-1α, IL-1β, IL-2Rα, IL-5, IL-6, IL-7, IL-8, G-CSF, GROα (CXCL1), MIP-1β (CCL4), MIG (CXCL9), RANTES (CCL5) and TRAIL in their BAL fluids. The only inflammatory markers that distinguished controlled asthma from uncontrolled asthma were neutrophil percentage and IL-8 levels, and both were inversely correlated with FEV₁. We examined whether grouping asthma subjects on the basis of BAL eosinophil % or neutrophil % could identify specific cytokine profiles. The only differences between neutrophil-normal asthma (neutrophil≤2.4%) and neutrophil-high asthma (neutrophils%>2.4%) were a higher BAL fluid IL-8 levels, and a lower FEV₁ in the latter group. By contrast, compared to eosinophil-normal asthma (eosinophils≤0.3%), eosinophil-high asthma (eosinophils>0.3%) had higher levels of IL-5, IL-13, IL-16, and PDGF-bb, but same neutrophil percentage, IL-8, and FEV₁. Our results identify neutrophils and IL-8 are the only inflammatory components in BAL fluids that distinguish controlled asthma from uncontrolled asthma, and both correlate inversely with FEV₁.     

DNA Methylation in Cosmc Promoter Region and Aberrantly Glycosylated IgA1 Associated with Pediatric IgA Nephropathy

IgA nephropathy (IgAN) is one of the most common glomerular diseases leading to end-stage renal failure. Elevation of aberrantly glycosylated IgA1 is a key feature of it. The expression of the specific molecular chaperone of core1ß1, 3galactosyl transferase (Cosmc) is known to be reduced in IgAN. We aimed to investigate whether the methylation of CpG islands of Cosmc gene promoter region could act as a possible mechanism responsible for down-regulation of Cosmc and related higher secretion of aberrantly glycosylated IgA1in lymphocytes from children with IgA nephropathy. Three groups were included: IgAN children (n = 26), other renal diseases (n = 11) and healthy children (n = 13). B-lymphocytes were isolated and cultured, treated or not with IL-4 or 5-Aza-2’-deoxycytidine (AZA). The levels of DNA methylation of Cosmc promotor region were not significantly different between the lymphocytes of the three children populations (P = 0.113), but there were significant differences between IgAN lymphocytes and lymphocytes of the other two children populations after IL-4 (P<0.0001) or AZA (P<0.0001). Cosmc mRNA expression was low in IgAN lymphocytes compared to the other two groups (P<0.0001). The level of aberrantly glycosylated IgA1 was markedly higher in IgAN group compared to the other groups (P<0.0001). After treatment with IL-4, the levels of Cosmc DNA methylation and aberrantly glycosylated IgA1 in IgAN lymphocytes were remarkably higher than the other two groups (P<0.0001) with more markedly decreased Cosmc mRNA content (P<0.0001). After treatment with AZA, the levels in IgAN lymphocytes were decreased, but was still remarkably higher than the other two groups (P<0.0001), while Cosmc mRNA content in IgAN lymphocytes were more markedly increased than the other two groups (P<0.0001). The alteration of DNA methylation by IL-4 or AZA specifically correlates in IgAN lymphocytes with alterations in Cosmc mRNA expression and with the level of aberrantly glycosylated IgA1 (r = −0.948, r = 0. 707). Our results suggested that hypermethylation of Cosmc promoter region could be a key mechanism for the reduction of Cosmc mRNA expression in IgAN lymphocytes with associated increase in aberrantly glycosylated IgA1.     

Targeted Next-Generation Sequencing in Uyghur Families with Non-Syndromic Sensorineural Hearing Loss

The mutation spectrum of deafness genes may vary in different ethnical groups. In this study, we investigated the genetic etiology of nonsyndromic deafness in four consanguineous and two multiplex Uyghur families in which mutations in common deafness genes GJB2, SLC26A4 and MT-RNR1 were excluded. Targeted next-generation sequencing of 97 deafness genes was performed in the probands of each family. Novel pathogenic mutations were identified in four probands including the p.L416R/p.A438T compound heterozygous mutations in TMC1, the homozygous p.V1880E mutation in MYO7A, c.1238delT frameshifting deletion in PCDH15 and c.9690+1G>A splice site mutation in MYO15A. Co-segregation of the mutations and the deafness were confirmed within each family by Sanger sequencing. No pathogenic mutations were identified in one multiplex family and one consanguineous family. Our study provided a useful piece of information for the genetic etiology of deafness in Uyghurs.     

Rapid Identification of α-Glucosidase Inhibitors from Phlomis tuberosa by Sepbox Chromatography and Thin-Layer Chromatography Bioautography

Alpha-glucosidase inhibitors currently form an important basis for developing novel drugs for diabetes treatment. In our preliminary tests, the ethyl acetate fraction of Phlomis tuberosa extracts showed significant α-glucosidase inhibitory activity (IC₅₀ = 100 μg/mL). In the present study, a combined method using Sepbox chromatography and thin-layer chromatography (TLC) bioautography was developed to probe α-glucosidase inhibitors further. The ethyl acetate fraction of P. tuberosa extracts was separated into 150 individual subfractions within 20 h using Sepbox chromatography. Then, under the guidance of TLC bioautography, 20 compounds were successfully isolated from these fractions, including four new diterpenoids [14-hydroxyabieta-8,11,13-triene-11-carbaldehyde-18-oic-12-carboxy-13-(1-hydroxy-1-methylethyl)-lactone (1), 14-hydroxyabieta-8,11,13-triene-17-oic-12-carboxy-13-(1-hydroxy-1-methylethyl)-lactone (2), 14,16-dihydroxyabieta-8,11,13-triene-15,17-dioic acid (3), and phlomisol (15,16-eposy-8,13(16),14-labdatrien-19-ol) (4)], and 16 known compounds. Activity estimation indicated that 15 compounds showed more potent α-glucosidase inhibitory effects (with IC₅₀ values in the range 0.067–1.203 mM) than the positive control, acarbose (IC₅₀ = 3.72 ± 0.113 mM). This is the first report of separation of α-glucosidase inhibitors from P. tuberosa.     

Chinese Social Media Reaction to Information about 42 Notifiable Infectious Diseases

This study aimed to identify what information triggered social media users’ responses regarding infectious diseases. Chinese microblogs in 2012 regarding 42 infectious diseases were obtained through a keyword search in the Weiboscope database. Qualitative content analysis was performed for the posts pertinent to each keyword of the day of the year with the highest daily count. Similar posts were grouped and coded. We identified five categories of information that increased microblog traffic pertaining to infectious diseases: news of an outbreak or a case; health education / information; alternative health information / Traditional Chinese Medicine; commercial advertisement / entertainment; and social issues. News unrelated to the specified infectious diseases also led to elevated microblog traffic. Our study showcases the diverse contexts from which increased social media traffic occur. Our results will facilitate better health communication as causes underlying increased social media traffic are revealed.