Is Bagging Effective in the Classification of Small-Sample Genomic and Proteomic Data?

There has been considerable interest recently in the application of bagging in the classification of both gene-expression data and protein-abundance mass spectrometry data. The approach is often justified by the improvement it produces on the performance of unstable, overfitting classification rules under small-sample situations. However, the question of real practical interest is whether the ensemble scheme will improve performance of those classifiers sufficiently to beat the performance of single stable, nonoverfitting classifiers, in the case of small-sample genomic and proteomic data sets. To investigate that question, we conducted a detailed empirical study, using publicly-available data sets from published genomic and proteomic studies. We observed that, under t-test and RELIEF filter-based feature selection, bagging generally does a good job of improving the performance of unstable, overfitting classifiers, such as CART decision trees and neural networks, but that improvement was not sufficient to beat the performance of single stable, nonoverfitting classifiers, such as diagonal and plain linear discriminant analysis, or 3-nearest neighbors. Furthermore, as expected, the ensemble method did not improve the performance of these classifiers significantly. Representative experimental results are presented and discussed in this work.     

Effects of carbon dioxide exposure on early brain development in rats

The developing brain is vulnerable to environmental factors. We investigated the effects of air that contained 0.05, 0.1 and 0.3% CO₂ on the hippocampus, prefrontal cortex (PFC) and amygdala. We focused on the circuitry involved in the neurobiology of anxiety, spatial learning, memory, and on insulin-like growth factor-1 (IGF-1), which is known to play a role in early brain development in rats. Spatial learning and memory were impaired by exposure to 0.3% CO₂ air, while exposure to 0.1 and 0.3% CO₂ air elevated blood corticosterone levels, intensified anxiety behavior, increased superoxide dismutase (SOD) enzyme activity and MDA levels in hippocampus and PFC; glutathione peroxidase (GPx) enzyme activity decreased in the PFC with no associated change in the hippocampus. IGF-1 levels were decreased in the blood, PFC and hippocampus by exposure to both 0.1 and 0.3% CO₂. In addition, apoptosis was increased, while cell numbers were decreased in the CA1 regions of hippocampus and PFC after 0.3% CO₂ air exposure in adolescent rats. A positive correlation was found between the blood IGF-1 level and apoptosis in the PFC. We found that chronic exposure to 0.3% CO₂ air decreased IGF-1 levels in the serum, hippocampus and PFC, and increased oxidative stress. These findings were associated with increased anxiety behavior, and impaired memory and learning.     

Direct effects of osteoprotegerin on human bone cell metabolism

Purpose

Osteoprotegerin (OPG) affects bone metabolism by intercepting the RANK-RANKL interaction which prevents osteoclastic differentiation and consequently reduces bone resorption. Different bone phenotypes of mice overexpressing OPG and of mice with knockdown of receptor activator of NF-κB (RANK) or RANK-ligand (RANKL) suggest that the mechanism of action of the OPG-RANKL-RANK system in regulating bone remodeling is not completely understood. Furthermore, OPG increases bone mass and density independently from reduced osteoclastogenesis which is consistent with the possibility that OPG may directly affect bone metabolism beyond its known role as decoy receptor for RANKL.

Methods

We treated primary human osteoblastic cells with OPG and inhibitory anti-RANKL antibodies and measured cellular ALP activity, in vitro mineralization, vitronectin receptor protein expression and ERK phosphorylation. We also analyzed the mRNA co-expression of ALP and OPG ex vivo in bone biopsies from acute and old stable vertebral fractures.

Results

OPG directly increased ALP activity and in vitro mineralization of HOC, enhanced expression of the vitronectin receptor thereby increasing adherence of HOC to vitronectin and stimulated ERK phosphorylation. All OPG-mediated effects could be prevented by RANKL antibodies or RANKL-siRNA transfection and MAPK inhibitor PD98059 reduced the stimulatory effect of OPG on integrin αv expression. In acutely fractured vertebrae OPG and ALP mRNA expression was significantly increased compared to stable vertebral fractures. In conclusion, OPG exerts direct osteoanabolic effects on HOC metabolism via RANKL in addition to its well described role as decoy receptor for RANKL.     

Vascular endothelial growth factor (VEGF)-induced up-regulation of CCN1 in osteoblasts mediates proangiogenic activities in endothelial cells and promotes fracture healing

Angiogenesis is indispensable during fracture repair, and vascular endothelial growth factor (VEGF) is critical in this process. CCN1 (CYR61) is an extracellular matrix signaling molecule that has been implicated in neovascularization through its interactions with several endothelial integrin receptors. CCN1 has been shown to be up-regulated during the reparative phase of fracture healing; however, the role of CCN1 therein remains unclear. Here, the regulation of CCN1 expression in osteoblasts and the functional consequences thereof were studied. Stimulation of osteoblasts with VEGF resulted in a dose- and time-dependent up-regulation of CCN1 mRNA and protein. An up-regulation of both cell surface-associated CCN1 as well as extracellular matrix-associated CCN1 in osteoblasts was found. The supernatant of VEGF-prestimulated osteoblasts was chemotactic for endothelial cells, increasing their migration and stimulated capillary-like sprout formation. These effects could be attributed to the presence of CCN1 in the osteoblast supernatant as they were prevented by an antibody against CCN1 or by small interfering RNA-mediated knockdown of osteoblast CCN1. Moreover, the supernatant of VEGF-prestimulated osteoblasts induced angiogenesis in Matrigel plugs in vivo in a CCN1-dependent manner. In addition, blockade of CCN1 prevented bone fracture healing in mice. Taken together, the present work demonstrates a potential paracrine loop consisting of the VEGF-mediated up-regulation of CCN1 in osteoblasts that attracts endothelial cells and promotes angiogenesis. Such a loop could be operative during fracture healing.     

Reliability of the calculated maximal lactate steady state in amateur cyclists

Complex performance diagnostics in sports medicine should contain maximal aerobic and maximal anaerobic performance. The requirements on appropriate stress protocols are high. To validate a test protocol quality criteria like objectivity and reliability are necessary. Therefore, the present study was performed in intention to analyze the reliability of maximal lactate production rate (V.L_amax) by using a sprint test, maximum oxygen consumption (V.O_2max) by using a ramp test and, based on these data, resulting power in calculated maximum lactate-steady-state (PMLSS) especially for amateur cyclists. All subjects (n = 23, age 26 ± 4 years) were leisure cyclists. At three different days they completed first a sprint test to approximate V.La_max. After 60 min of recreation time a ramp test to assess V.O_2max was performed. The results of V.La_max-test and V.O_2max-test and the body weight were used to calculate PMLSS for all subjects. The intra class correlation (ICC) for V.La_max and V.O_2max was 0.904 and 0.987, respectively, coefficient of variation (CV) was 6.3% and 2.1%, respectively. Between the measurements the reliable change index of 0.11 mmol·l ^-1s ^-1 for V.La_max and 3.3 mlkg ^-1min ^-1 for V.O_2max achieved significance. The mean of the calculated PMLSS was 237 ± 72 W with an RCI of 9 W and reached with ICC = 0.985 a very high reliability. Both metabolic performance tests and the calculated PMLSS are reliable for leisure cyclists.     

Application of confocal laser scanning microscopy to detect oxidative stress in human colon cells

Introduction Excess of intracellular reactive oxygen species in relation to antioxidative systems results in an oxidative environment which may modulate gene expression or damage cellular molecules. These events are expected to greatly contribute to processes of carcinogenesis. Only few studies are available on the oxidative/reductive conditions in the colon, an important tumour target tissue. It was the objective of this work to further develop methods to assess intracellular oxidative stress within human colon cells as a tool to study such associations in nutritional toxicology.

Methods We have measured H₂O₂-induced oxidative stress in different colon cell lines, in freshly isolated human colon crypts, and, for comparative purposes, in NIH3T3 mouse embryo fibroblasts. Detection was performed by loading the cells with the fluorigenic peroxide-sensitive dye 6-carboxy-2′,7′-dichlorodihydrofluorescein diacetate (diacetoxymethyl ester), followed by in vitro treatment with H₂O₂ and fluorescence detection with confocal laser scanning microscopy (CLSM). Using the microgel electrophoresis (“Comet”) Assay, we also examined HT29 stem and clone 19A cells and freshly isolated primary colon cells for their relative sensitivity toward H₂O₂-induced DNA damage and for steady-state levels of endogenous oxidative DNA damage.

Results A dose-response relationship was found for the H₂O₂-induced dye decomposition in NIH3T3 cells (7.8–125 μM H₂O₂) whereas no effect occurred in the human colon tumour cell lines HT29 stem and HT29 clone 19A (62–1000 μM H₂O₂). Fluorescence was significantly increased at 62 μM H₂O₂ in the human colon adenocarcinoma cell line Caco-2. In isolated human colon crypts, the lower crypt cells (targets of colon cancer) were more sensitive towards H₂O₂ than the more differentiated upper crypt cells. In contrast to the CLSM results, oxidative DNA damage was detected in both cell lines using the Comet Assay. Endogenous oxidative DNA damage was highest in HT29 clone 19A, followed by the primary colon cells and HT29 stem cells.

Conclusions Oxidative stress in colon cells leads to damage of macromolecules which is sensitively detected in the Comet Assay. The lacking response of the CLSM-approach in colon tumour cells is probably due to intrinsic modes of protective activities of these cells. In general, however, the CLSM method is a sensitive technique to detect very low concentrations of H₂O₂-induced oxidative stress in NIH3T3 cells. Moreover, by using colon crypts it provides the unique possibility of assessing cell specific levels of oxidative stress in explanted human tissues. Our results demonstrate that the actual target cells of colon cancer induction are indeed susceptible to the oxidative activity of H₂O₂.     

Fruit juice consumption modulates antioxidative status,immune status and DNA damage

Polyphenolic compounds exert a variety of physiological effects in vitro including antioxidative, immunomodulatory and antigenotoxic effects. In a randomized crossover study in healthy men on a low-polyphenol diet, we determined the effects of 2 polyphenol-rich juices (330 ml/d) supplemented for 2 weeks on bioavailability of polyphenols, markers of antioxidative and immune status, and reduction of DNA damage. Juices provided 236 mg (A) and 226 mg (B) polyphenols with cyanidin glycosides (A) and epigallocatechin gallate (B) as major polyphenolic ingredients. There was no accumulation of plasma polyphenols after two weeks of juice supplementation. In contrast, plasma malondialdehyde decreased with time during juice interventions. Moreover, juice consumption also increased lymphocyte proliferative responsiveness, with no difference between the two juices. Interleukin-2 secretion by activated lymphocytes and the lytic activity of natural killer cells were significantly increased by both juices. Juice intervention had no effect on single DNA strand breaks, but significantly reduced oxidative DNA damage in lymphocytes. A time-delay was observed between the intake of fruit juice and the reduction of oxidative DNA damage and the increase in interleukin-2 secretion. We conclude that consumption of either juice enhanced antioxidant status, reduced oxidative DNA damage and stimulated immune cell functions. However, fruit juice consumption for 2 weeks did not result in elevated plasma polyphenols in subjects after overnight fasting. Further studies should focus on the time-delay between juice intake and changes in measured physiological functions, as well as on active polyphenolic metabolites mediating the observed effects.     

Genetic Stability of Foamy Viruses: Long-Term Study in an African Green Monkey Population

The genetic variability of the envelope surface domain (SU) of simian foamy virus (FV) of African green monkeys was studied. To assess the interindividual diversity of FV, isolates were obtained from 19 animals living together in a monkey house. The monkeys had been imported from Kenya prior to being placed in long-term housing in the research institute. In addition, a simian FV isolate and proviral DNA were obtained from an animal caretaker infected in this setting. DNA of the complete SU (1779 to 1793 bp) was analyzed by PCR and sequencing. The sequences revealed four clusters with high homologies (>95%). Between the clusters, divergencies ranged from 3 to 25%. Obviously, the clusters reflect four different strains or subtypes of simian FV type 3 that were prevalent in the colony. In contrast to lentiviruses, hypervariable regions could not be detected in the FV SU. Furthermore, to analyze the intraindividual diversity of FV, we investigated the virus population within an individual monkey at a given time point and its evolution over 13 years. For this purpose, 22 proviral SU clones generated by PCR from one oral swab and seven isolates obtained from the same animal between 1982 and 1995 were examined. These sequences revealed exceptionally high homology rates (99.5 to 100%), and only a minimal genetic drift was recognized within the series of isolates. In conclusion, the low in vivo divergency of FV SU suggests that genetic variability is not important for the maintenance of FV persistence.     

Mutagenic activity of carcinogens detected in transgenic rodent mutagenicity assays at dose levels used in chronic rodent cancer bioassays

Data on transgenic rodent mutagenicity of five human carcinogens were summarised and compared with the results from rodent carcinogenicity studies. Four out of five carcinogens showed mutagenic activity already at daily dose levels which induced cancer in long-term rodent bioassays in at least one target tissue of carcinogenesis. In several of these studies, even single dose applications were sufficient to significantly increase the mutation frequency in vivo. Other genotoxic carcinogens required application of multiple dosing at dose-levels used in rodent cancer bioassays to show their in vivo mutagenicity. A rodent respiratory tract carcinogen, 1,2-dibromoethane (DBE), following inhalation exposure, displayed no mutagenic activity, neither in lung nor in nasal mucosa, at a single 2-h exposure to 30 ppm, which is below the highest concentration used in a NTP cancer bioassay. In contrast, after multiple treatment for 10 days at the same daily doses, a significant increase of the mutation frequency in nasal mucosa was apparent. We conclude, that especially when studying new chemicals in these transgenic rodent mutation assays, a multiple dosing protocol should be preferred. For dose selection, the same criteria could be applied as for chronic rodent bioassays.     

Androgen receptor isoforms AR-A and AR-B display functional differences in cultured human bone cells and genital skin fibroblasts

Two isoforms of the androgen receptor (AR-A and AR-B), differing by a lack of the first 187 amino acids in the NH2-terminal transactivation domain of AR-A, are expressed in connective tissue and bone. Transient transfections of normal human osteoblastic cells (HOB) and of genital skin fibroblasts defective in AR (GSF-540) were utilized to compare the functional properties of AR isoforms in mesenchymal tissues. Overexpression of AR-B or AR-A did not significantly affect type I collagen secretion. However, overexpression of AR-B (but not AR-A) restored androgen-dependent DNA synthesis in AR-defective fibroblasts and increased DHT-mediated DNA synthesis three-fold in osteoblastic cells. Overexpression of AR-A did not affect DHT action but reduced DHT-dependent DNA synthesis when transfected together with AR-B. The need for an NH2-terminal sequence of the AR for complete receptor function was demonstrated using electrophoretic mobility shift assay. A peptide coding for the amino terminus of the complete AR was able to decrease the binding affinity of AR-B and increase the binding affinity of AR-A to the androgen response element. Our results suggest that AR-A lacks the ability to stimulate cell proliferation possibly due to reduced binding of AR co-activating proteins to the truncated N-terminal transactivation domain rather than due to impaired stability of the AR-A isoform.