A comprehensive sensitivity analysis of microarray breast cancer classification under feature variability

Background  

Large discrepancies in signature composition and outcome concordance have been observed between different microarray breast cancer expression profiling studies. This is often ascribed to differences in array platform as well as biological variability. We conjecture that other reasons for the observed discrepancies are the measurement error associated with each feature and the choice of preprocessing method. Microarray data are known to be subject to technical variation and the confidence intervals around individual point estimates of expression levels can be wide. Furthermore, the estimated expression values also vary depending on the selected preprocessing scheme. In microarray breast cancer classification studies, however, these two forms of feature variability are almost always ignored and hence their exact role is unclear.     

Differential Specificity and Immunogenicity of Adenovirus Type 5 Neutralizing Antibodies Elicited by Natural Infection or Immunization

A recent clinical trial of a T-cell-based AIDS vaccine delivered with recombinant adenovirus type 5 (rAd5) vectors showed no efficacy in lowering viral load and was associated with increased risk of human immunodeficiency virus type 1 (HIV-1) infection. Preexisting immunity to Ad5 in humans could therefore affect both immunogenicity and vaccine efficacy. We hypothesized that vaccine-induced immunity is differentially affected, depending on whether subjects were exposed to Ad5 by natural infection or by vaccination. Serum samples from vaccine trial subjects receiving a DNA/rAd5 AIDS vaccine with or without prior immunity to Ad5 were examined for the specificity of their Ad5 neutralizing antibodies and their effect on HIV-1 immune responses. Here, we report that rAd5 neutralizing antibodies were directed to different components of the virion, depending on whether they were elicited by natural infection or vaccination in HIV vaccine trial subjects. Neutralizing antibodies elicited by natural infection were directed largely to the Ad5 fiber, while exposure to rAd5 through vaccination elicited antibodies primarily to capsid proteins other than fiber. Notably, preexisting immunity to Ad5 fiber from natural infection significantly reduced the CD4 and CD8 cell responses to HIV Gag after DNA/rAd5 vaccination. The specificity of Ad5 neutralizing antibodies therefore differs depending on the route of exposure, and natural Ad5 infection compromises Ad5 vaccine-induced immunity to weak immunogens, such as HIV-1 Gag. These results have implications for future AIDS vaccine trials and the design of next-generation gene-based vaccine vectors.Recombinant adenovirus (rAd)-based vectors are currently under investigation in a variety of gene therapy and T-cell-based vaccine clinical trials. There are more than 370 such ongoing clinical trials for broad applications, including infectious diseases and cancer therapy (http://www.wiley.co.uk/genetherapy/clinical/). Based on supportive data from nonhuman primate studies, rAd-based vectors have been developed and tested in human clinical trials to deliver human immunodeficiency virus (HIV-1) gene products that stimulate HIV-specific immune responses. Preexisting immunity to Ad serotype 5 (Ad5), from which most vectors are derived, is common in humans. Though neutralizing antibodies to Ad5 may reduce the immunogenicity of Ad5-based vectors in animal models (16), their effect on immunity in subjects with previous Ad5 infection is poorly understood. In the STEP trial, which tested a Merck rAd5 vaccine encoding HIV-1 Gag, Pol, and Nef, vaccination failed to show protection, either by lowering viral load or by decreasing acquisition of infection (3, 9, 12, 21). Furthermore, the possibility was raised that subjects with preexisting neutralizing antibodies from natural Ad5 infection may have carried an increased risk of HIV infection after vaccination. Thus, understanding the nature and immune effects of Ad5 seropositivity in humans is important to the development of vaccines against AIDS and other diseases.Ad5 is a common cause of respiratory disease and an occasional cause of gastroenteritis in humans, and exposure before adolescence is common in human populations (19). Such exposure stimulates both innate and adaptive immune responses that generate neutralizing antibodies and virus-specific T-cell responses (6). These antibodies can also synergize with each other to achieve maximum viral neutralization (7, 22). The capsid protein specificity of Ad5 neutralizing antibodies has been reported for humans following administration of rAd5 gene therapy vectors for advanced liver or lung cancer (7, 10). However, results were presented solely for antibodies induced by administration of rAd5. One report has assessed Ad5 neutralizing antibodies with a healthy human population that was Ad5 seropositive from natural exposure to the virus (18). The median titer of the population was presented, but the frequency of protein-specific neutralizing antibody has not been defined for humans.Here we describe the first report of the natural frequency and effect on immunization of neutralizing antibodies specific for different Ad capsid proteins in human subjects. We address the fundamental mechanisms of how humans generate neutralizing antibodies to a common cold virus that is in widespread use as a vector for gene therapy and vaccines. Such mechanisms may also be applicable to other nonenveloped viruses, including adeno-associated viruses and other viruses containing multiple envelope surface proteins, like influenza. To analyze the contribution of anti-capsid antibodies to neutralization by different human serum samples, wild-type and chimeric vectors were utilized. For example, a rAd type 5 (rAd5) vector with a fiber derived from Ad35 fiber (rAd5 F35) can be used to analyze the anti-Ad5 capsid response independent of fiber. Conversely, a rAd35 vector with a fiber transposed from Ad5 can determine the specificity of neutralization mediated by the Ad5 fiber. Using these vectors, we have analyzed human serum samples from two HIV vaccine clinical trials, VRC 006 and HVTN 204, in which a single-dose rAd5 vaccine alone and a three-dose DNA prime/single dose rAd5 boost vaccine encoding HIV-1 Env A,B, and C; Gag; and Pol, respectively, were administered. Thus, we sought to characterize the specificity of rAd5 neutralizing antibodies in Ad5-immune subjects and to determine their effect on immune responses elicited by vaccination.     

HIV-DNA priming alters T cell responses to HIV-adenovirus vaccine even when responses to DNA are undetectable

Many candidate HIV vaccines are designed to primarily elicit T cell responses. Although repeated immunization with the same vaccine boosts Ab responses, the benefit for T cell responses is ill defined. We compared two immunization regimens that include the same recombinant adenoviral serotype 5 (rAd5) boost. Repeated homologous rAd5 immunization fails to increase T cell responses, but increases gp140 Ab responses 10-fold. DNA prime, as compared with rAd5 prime, directs long-term memory CD8(+) T cells toward a terminally differentiated effector memory phenotype with cytotoxic potential. Based on the kinetics of activated cells measured directly ex vivo, the DNA vaccination primes for both CD4(+) and CD8(+) T cells, despite the lack of detection of the latter until after the boost. These results suggest that heterologous prime-boost combinations have distinct immunological advantages over homologous prime-boosts and suggest that the effect of DNA on subsequent boosting may not be easily detectable directly after the DNA vaccination.     

Demographic processes affect HIV-1 evolution in primary infection before the onset of selective processes

HIV-1 transmission and viral evolution in the first year of infection were studied in 11 individuals representing four transmitter-recipient pairs and three independent seroconverters. Nine of these individuals were enrolled during acute infection; all were men who have sex with men (MSM) infected with HIV-1 subtype B. A total of 475 nearly full-length HIV-1 genome sequences were generated, representing on average 10 genomes per specimen at 2 to 12 visits over the first year of infection. Single founding variants with nearly homogeneous viral populations were detected in eight of the nine individuals who were enrolled during acute HIV-1 infection. Restriction to a single founder variant was not due to a lack of diversity in the transmitter as homogeneous populations were found in recipients from transmitters with chronic infection. Mutational patterns indicative of rapid viral population growth dominated during the first 5 weeks of infection and included a slight contraction of viral genetic diversity over the first 20 to 40 days. Subsequently, selection dominated, most markedly in env and nef. Mutants were detected in the first week and became consensus as early as day 21 after the onset of symptoms of primary HIV infection. We found multiple indications of cytotoxic T lymphocyte (CTL) escape mutations while reversions appeared limited. Putative escape mutations were often rapidly replaced with mutually exclusive mutations nearby, indicating the existence of a maturational escape process, possibly in adaptation to viral fitness constraints or to immune responses against new variants. We showed that establishment of HIV-1 infection is likely due to a biological mechanism that restricts transmission rather than to early adaptive evolution during acute infection. Furthermore, the diversity of HIV strains coupled with complex and individual-specific patterns of CTL escape did not reveal shared sequence characteristics of acute infection that could be harnessed for vaccine design.     

Duodenal mucosal risk markers in patients with familial adenomatous polyposis: effects of celecoxib/ursodeoxycholic acid co-treatment and comparison with patient controls

Background

Familial adenomatous polyposis (FAP) is a disease characterized by the development of hundreds to thousands of adenomatous polyps in the colorectum early in life. Virtually all patients with FAP will develop colorectal cancer before the age of 40 to 50 years, unless prophylactic colectomy is performed, which significantly improves their prognosis. The mortality pattern has changed and duodenal cancer now is one of the main cancer-related causes of death in these patients. Practically all patients with FAP develop premalignant duodenal adenomas, which may develop to duodenal cancer in approximately 3-7% of patients. Duodenal cancer in patients with FAP has a poor prognosis. The clinical challenge is to identify patients at high-risk for duodenal carcinoma. Chemoprevention would be desirable to avoid duodenectomy. The main goal of this study is to identify risk markers in normal duodenal mucosa of patients with FAP, that could help identify patients at increased risk for malignant transformation.

Methods

Messenger RNA (mRNA) levels of glutathione S-transferase A1 (GSTA1), glutathione S-transferase P1 (GSTP1), KIAA1199, E-cadherin, peroxisome proliferative activated receptor δ (PPARδ), caspase-3, cyclin D1, β-catenin, and cyclooxygenase-2 (COX-2) were measured in duodenal mucosa, using the QuantiGene 2.0 Plex assay. Levels in normal appearing mucosa of patients with FAP (n?=?37) were compared with levels in non-FAP patient controls (n?=?16). In addition, levels before and after treatment with either celecoxib & ursodeoxycholic acid (UDCA, n?=?14) or celecoxib & placebo (n?=?13) were evaluated in patients with FAP.

Results

mRNA levels of glutathione S-transferase A1 (28.16% vs. 38.24%, p?=?0.008) and caspase-3 (3.30% vs. 5.31%, p?=?0.001) were significantly lower in patients with FAP vs. non-FAP patient controls, respectively. COX-2 mRNA levels in normal duodenal mucosa of patients with FAP were found to be unexpectedly low. None of the potential risk markers was influenced by celecoxib or celecoxib & UDCA.

Conclusions

Protection against toxins and carcinogens (GSTA1) and apoptosis (caspase-3) is low in patients with FAP, which could contribute to increased susceptibility for malignant transformation of duodenal mucosa.

Trial registration

http://ClinicalTrials.gov number NCT00808743

     

Inflammation-related proteins in the blood of extremely low gestational age newborns. The contribution of inflammation to the appearance of developmental regulation

BackgroundWe wanted to assess to what extent concentrations of circulating proteins appear to be developmentally regulated, and to what extent such regulation is influenced by intra-uterine inflammation.MethodsWe measured 22 proteins in blood obtained on postnatal days 1, 7, and 14 from 818 children born before the 28th week of gestation for whom we also had information about placenta morphology.ResultsWithin the narrow gestational age range of this sample, some protein concentrations increase in blood with increasing gestational age. More commonly, the concentrations of inflammation-related proteins decrease with increasing gestational age. We observed this inverse pattern both in children whose placenta was and was not inflamed.Conclusions/InferencesRegardless of whether or not the placenta is inflamed, the concentrations of inflammation-related proteins in early blood specimens appear to be developmentally regulated with the most common pattern being a decrease with increasing gestational age.     

Mitochondrial mismatch analysis is insensitive to the mutational process

Mismatch distributions are histograms showing the pattern of nucleotide (or restriction) site differences between pairs of individuals in a sample. They can be used to test hypotheses about the history of population size and subdivision (if selective neutrality is assumed) or about selection (if a constant population size is assumed). Previous work has assumed that mutations never strike the same site twice, an assumption that is called the model of infinite sites. Fortunately, the results are surprisingly robust even when this assumption is violated. We show here that (1) confidence regions inferred using the infinite- sites model differ little from those inferred using a model of finite sites with uniform site-specific mutation rates, and (2) even when site- specific mutation rates follow a gamma distribution, confidence regions are little changed until the gamma shape parameter falls well below its plausible range, to roughly 0.01. In addition, we evaluate and reject the proposition that mismatch waves are produced by pooling data from several subdivisions of a structured population.