Search

Automated variable selection procedures, such as backward elimination, are commonly employed to perform model selection in the context of multivariable regression. The stability of such procedures can be investigated using a bootstrap-based approach. The idea is to apply the variable selection procedure on a large number of bootstrap samples successively and to examine the obtained models, for instance, in terms of the inclusion of specific predictor variables. In this paper, we aim to investigate a particular important problem affecting this method in the case of categorical predictor variables with different numbers of categories and to give recommendations on how to avoid it. For this purpose, we systematically assess the behavior of automated variable selection based on the likelihood ratio test using either bootstrap samples drawn with replacement or subsamples drawn without replacement from the original dataset. Our study consists of extensive simulations and a real data example from the NHANES study. Our main result is that if automated variable selection is conducted on bootstrap samples, variables with more categories are substantially favored over variables with fewer categories and over metric variables even if none of them have any effect. Importantly, variables with no effect and many categories may be (wrongly) preferred to variables with an effect but few categories. We suggest the use of subsamples instead of bootstrap samples to bypass these drawbacks., (© 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim.)

Background/aims: Theoretically, the trait-model parameters (disease allele frequency and penetrance function) can be estimated without bias in a MOD score linkage analysis. We aimed to practically evaluate the MOD score approach regarding its ability to provide unbiased trait-model parameters for various pedigree-type and trait-model scenarios. We further investigated the ability of the MOD score approach to detect imprinting using affected sib pairs (ASPs) and affected half-sib pairs (AHSPs) when all parental genotypes are missing., Methods: Simulated pedigree data were analyzed using the GENEHUNTER-MODSCORE software package. Parameter estimation performance in terms of bias and variability was evaluated with regard to trait-model type and pedigree complexity., Results: Generally, parameters were estimated with lower bias and variability with increasing pedigree complexity, especially for recessive and overdominant models. However, dominant and additive models could hardly be distinguished even when using 3-generation pedigrees. Imprinting could clearly be detected for mixtures of mainly ASPs and only few AHSPs with the common parent of the imprinted sex, even though no parental genotypes were available., Conclusion: Our results provide guidance to researchers regarding the possibility to estimate trait-model parameters by a MOD score analysis, including the degree of imprinting, with certain types of pedigrees., (© 2017 The Author(s) Published by S. Karger AG, Basel.)

It has been suggested that mitochondrial dysfunction has an influence on lipid metabolism. The fact that mitochondrial defects can be accumulated over time as a normal part of aging may explain why cholesterol levels often are altered with age. To test the hypothesis whether mitochondrial variants are associated with lipid profile (total cholesterol, LDL, HDL, and triglycerides) we analyzed a total number of 978 mitochondrial single nucleotide polymorphisms (mtSNPs) in a sample of 2,815 individuals participating in the population-based KORA F4 study. To assess mtSNP association while taking the presence of heteroplasmy into account we used the raw signal intensity values measured on the microarray and applied linear regression. Ten mtSNPs (mt3285, mt3336, mt5285, mt6591, mt6671, mt9163, mt13855, mt13958, mt14000, and mt14580) were significantly associated with HDL cholesterol and one mtSNP (mt15074) with triglycerides levels. These results highlight the importance of the mitochondrial genome among the factors that contribute to the regulation of lipid levels. Focusing on mitochondrial variants may lead to further insights regarding the underlying physiological mechanisms, or even to the development of innovative treatments. Since this is the first mitochondrial genome-wide association analysis (mtGWAS) for lipid profile, further analyses are needed to follow up on the present findings.

Screening programs are recommended for individuals at risk (IAR) from families with familial pancreatic cancer (FPC). However, reliable imaging methods or biomarkers for early diagnosis of pancreatic ductal adenocarcinoma (PC) or its precursor lesions are still lacking. The ability of circulating microRNAs (miRNAs) to discriminate multifocal high-grade precursor lesions or PC from normal was examined. The presence of miRNA-21, -155, -196a, -196b and -210 was analyzed in the serum of transgenic KPC mice to test their ability to distinguish mice with different grades of pancreatic intraepithelial neoplasia (mPanIN1-3) or PC from control mice. Serum levels of miR-196a and -196b were significantly higher in mice with PanIN2/3 lesions (n = 10) or PC (n = 8) as compared to control mice (n = 10) or mice with PanIN1 lesions (n = 10; P = .01). In humans, miR-196a and -196b were also diagnostic. Patients with PC, sporadic (n = 9) or hereditary (n = 10), and IAR with multifocal PanIN2/3 lesions (n = 5) had significantly higher serum levels than patients with neuroendocrine pancreatic tumors (n = 10) or chronic pancreatitis (n = 10), IAR with PanIN1 or no PanIN lesions (n = 5), and healthy controls (n = 10). The combination of both miR-196a and -196b reached a sensitivity of 1 and specificity of 0.9 (area under the curve = 0.99) to diagnose PC or high-grade PanIN lesions. In addition, preoperative elevated serum levels of miR-196a and -196b in patients with PC or multifocal PanIN2/3 lesions dropped to normal after potential curative resection. The combination of miR-196a and -196b may be a promising biomarker test for the screening of IAR for FPC., (Copyright © 2014 The Authors. Published by Elsevier Inc. All rights reserved.)

It has been suggested that mitochondrial dysfunction plays a role in the pathogenesis of asthma. To test whether mitochondrial variants influence the risk of asthma, we analyzed 16,158 mtSNPs in a sample of 372 asthmatic children and 395 healthy children using the DNA pooling technique and genome wide association analysis. Stratified analysis by sex was performed to explain the differences observed between sexes in the etiology of asthma. Different variants were detected to be significant in the sample of girls and boys with the smallest adjusted p values being 1.4 × 10(-09) (mt5295) and 3.6 × 10(-12) (mt16158), respectively. Most of the significant locations found in boys are within the CYB gene and the non-coding region. For girls, most of the significant mtSNPs lie within NADH-dehydrogenase-subunits. The variants reported here have not previously been described in connection with asthma. Although further studies in other cohorts are needed to confirm these findings our study highlights the importance of the mitochondria among the factors that contribute to the risk of asthma., (Copyright © 2013 Elsevier B.V. All rights reserved.)

High-risk individuals of familial pancreatic cancer (FPC) families are considered to be good candidates for screening programs to detect early PC or its high-grade precursor lesions, especially pancreatic intraepithelial neoplasia (PanIN) 2/3 lesions. There is a definite need for diagnostic markers as neither reliable imaging methods nor biomarkers are available to detect these lesions. On the basis of a literature search, the potential serum markers neutrophil gelatinase-associated lipocalin (LCN2), metallopeptidase inhibitor 1 (TIMP1), chemokine (C-X-C motif) ligand 16 (CXCL16), IGFBP4, and iC3a, which were first tested in transgenic KrasLSL.(G12D/+);p53(R172H/+);Pdx1-Cre mice, were identified. ELISA analyses of LCN2, TIMP1, and CXCL16 revealed significantly higher levels in mice with PanIN2/3 lesions or PC compared to mice with normal pancreata or PanIN1 lesions. Analysis of preoperative human serum samples from patients with sporadic PC (n = 61), hereditary PC (n = 24), chronic pancreatitis (n = 28), pancreatic neuroendocrine tumors (n = 11), and FPC patients with histologically proven multifocal PanIN2/3 lesions (n = 3), as well as healthy control subjects (n = 20), confirmed significantly higher serum levels of LCN2 and TIMP1 in patients with PC and multifocal PanIN2/3 lesions. The combination of LCN2 and TIMP1 as a diagnostic test for the detection of PC had a sensitivity, specificity, and positive predictive value of 100% each. Although this preliminary finding needs to be validated in a large series of individuals at high risk for FPC, serum measurement of LCN2 and TIMP1 might be a promising screening tool.