Search

Recent advance in biotechnology and its wide applications have led to the generation of many high-dimensional gene expression data sets that can be used to address similar biological questions. Meta-analysis plays an important role in summarizing and synthesizing scientific evidence from multiple studies. When the dimensions of datasets are high, it is desirable to incorporate variable selection into meta-analysis to improve model interpretation and prediction. According to our knowledge, all existing methods conduct variable selection with meta-analyzed data in an "all-in-or-all-out" fashion, that is, a gene is either selected in all of studies or not selected in any study. However, due to data heterogeneity commonly exist in meta-analyzed data, including choices of biospecimens, study population, and measurement sensitivity, it is possible that a gene is important in some studies while unimportant in others. In this article, we propose a novel method called meta-lasso for variable selection with high-dimensional meta-analyzed data. Through a hierarchical decomposition on regression coefficients, our method not only borrows strength across multiple data sets to boost the power to identify important genes, but also keeps the selection flexibility among data sets to take into account data heterogeneity. We show that our method possesses the gene selection consistency, that is, when sample size of each data set is large, with high probability, our method can identify all important genes and remove all unimportant genes. Simulation studies demonstrate a good performance of our method. We applied our meta-lasso method to a meta-analysis of five cardiovascular studies. The analysis results are clinically meaningful., (© 2014, The International Biometric Society.)

Background & Aims: Levels of the thiopurine metabolites 6-thioguanine nucleotide (6-TGN) and 6-methylmercaptopurine commonly are monitored during thiopurine therapy for inflammatory bowel disease despite this test's high cost and poor prediction of clinical response (sensitivity, 62%; specificity, 72%). We investigated whether patterns in common laboratory parameters might be used to identify appropriate immunologic responses to thiopurine and whether they are more accurate than measurements of thiopurine metabolites in identifying patients who respond to therapy., Methods: We identified 774 patients with inflammatory bowel disease on thiopurine therapy using metabolite and standard laboratory tests over a 24-hour time period. Machine learning algorithms were developed using laboratory values and age in a random training set of 70% of the cases; these algorithms were tested in the remaining 30% of the cases., Results: A random forest algorithm was developed based on laboratory and age data; it differentiated clinical responders from nonresponders in the test set with an area under the receiver operating characteristic (AUROC) curve of 0.856. In contrast, 6-TGN levels differentiated clinical responders from nonresponders with an AUROC of 0.594 (P < .001). Algorithms developed to identify thiopurine nonadherence (AUROC, 0.813) and thiopurine shunters (AUROC, 0.797) were accurate., Conclusions: Algorithms that use age and laboratory values can differentiate clinical response, nonadherence, and shunting of thiopurine metabolism among patients who take thiopurines. This approach was less costly and more accurate than 6-TGN metabolite measurements in predicting clinical response. If validated, this approach would provide a low-cost, rapid alternative to metabolite measurements for monitoring thiopurine use., (Copyright 2010 AGA Institute. Published by Elsevier Inc. All rights reserved.)

Recent interest in cancer research focuses on predicting patients' survival by investigating gene expression profiles based on microarray analysis. We propose a doubly penalized Buckley-James method for the semiparametric accelerated failure time model to relate high-dimensional genomic data to censored survival outcomes, which uses the elastic-net penalty that is a mixture of L1- and L2-norm penalties. Similar to the elastic-net method for a linear regression model with uncensored data, the proposed method performs automatic gene selection and parameter estimation, where highly correlated genes are able to be selected (or removed) together. The two-dimensional tuning parameter is determined by generalized crossvalidation. The proposed method is evaluated by simulations and applied to the Michigan squamous cell lung carcinoma study.

Motivation: The nearest shrunken centroid (NSC) method has been successfully applied in many DNA-microarray classification problems. The NSC uses 'shrunken' centroids as prototypes for each class and identifies subsets of genes that best characterize each class. Classification is then made to the nearest (shrunken) centroid. The NSC is very easy to implement and very easy to interpret, however, it has drawbacks., Results: We show that the NSC method can be interpreted in the framework of LASSO regression. Based on that, we consider two new methods, adaptive L(infinity)-norm penalized NSC (ALP-NSC) and adaptive hierarchically penalized NSC (AHP-NSC), with two different penalty functions for microarray classification, which improve over the NSC. Unlike the L(1)-norm penalty used in LASSO, the penalty terms that we consider make use of the fact that parameters belonging to one gene should be treated as a natural group. Numerical results indicate that the two new methods tend to remove irrelevant genes more effectively and provide better classification results than the L(1)-norm approach., Availability: R code for the ALP-NSC and the AHP-NSC algorithms are available from authors upon request.

In this paper, we propose variable selection and estimation in generalized linear models using the seamless $L_0$ (SELO) penalized likelihood approach. The SELO penalty is a smooth function that very closely resembles the discontinuous $L_0$ penalty. We develop an efficient algorithm to fit the model, and show that the SELO-GLM procedure has the oracle property in the presence of a diverging number of variables. We propose a Bayesian information criterion (BIC) to select the tuning parameter. We show that under some regularity conditions, the proposed SELO-GLM/BIC procedure consistently selects the true model. We perform simulation studies to evaluate the finite sample performance of the proposed methods. Our simulation studies show that the proposed SELO-GLM procedure has a better finite sample performance than several existing methods, especially when the number of variables is large and the signals are weak. We apply the SELO-GLM to analyze a breast cancer genetic dataset to identify the SNPs that are associated with breast cancer risk. The Canadian Journal of Statistics 40: 745-769; 2012 © 2012 Statistical Society of Canada [ABSTRACT FROM AUTHOR]

Variable selection in high-dimensional clustering analysis is an important yet challenging problem. In this article, we propose two methods that simultaneously separate data points into similar clusters and select informative variables that contribute to the clustering. Our methods are in the framework of penalized model-based clustering. Unlike the classical L1-norm penalization, the penalty terms that we propose make use of the fact that parameters belonging to one variable should be treated as a natural “group.” Numerical results indicate that the two new methods tend to remove noninformative variables more effectively and provide better clustering results than the L1-norm approach. [ABSTRACT FROM AUTHOR]