Hierarchical Bayes models for cDNA microarray gene expression

cDNA microarrays are used in many contexts to compare mRNA levels between samples of cells. Microarray experiments typically give us expression measurements on 1000-20 000 genes, but with few replicates for each gene. Traditional methods using means and standard deviations to detect differential expression are not satisfactory in this context. A handful of alternative statistics have been developed, including several empirical Bayes methods. In the present paper we present two full hierarchical Bayes models for detecting gene expression, of which one (D) describes our microarray data very well. We also compare the full Bayes and empirical Bayes approaches with respect to model assumptions, false discovery rates and computer running time. The proposed models are compared to existing empirical Bayes models in a simulation study and for a set of data (Yuen et al., 2002), where 27 genes have been categorized by quantitative real-time PCR. It turns out that the existing empirical Bayes methods have at least as good performance as the full Bayes ones.     

Empirical Bayes screening of many p-values with applications to microarray studies

MOTIVATION: Statistical tests for the detection of differentially expressed genes lead to a large collection of p-values one for each gene comparison. Without any further adjustment, these p-values may lead to a large number of false positives, simply because the number of genes to be tested is huge, which might mean wastage of laboratory resources. To account for multiple hypotheses, these p-values are typically adjusted using a single step method or a step-down method in order to achieve an overall control of the error rate (the so-called familywise error rate). In many applications, this may lead to an overly conservative strategy leading to too few genes being flagged. RESULTS: In this paper we introduce a novel empirical Bayes screening (EBS) technique to inspect a large number of p-values in an effort to detect additional positive cases. In effect, each case borrows strength from an overall picture of the alternative hypotheses computed from all the p-values, while the entire procedure is calibrated by a step-down method so that the familywise error rate at the complete null hypothesis is still controlled. It is shown that the EBS has substantially higher sensitivity than the standard step-down approach for multiple comparison at the cost of a modest increase in the false discovery rate (FDR). The EBS procedure also compares favorably when compared with existing FDR control procedures for multiple testing. The EBS procedure is particularly useful in situations where it is important to identify all possible potentially positive cases which can be subjected to further confirmatory testing in order to eliminate the false positives. We illustrated this screening procedure using a data set on human colorectal cancer where we show that the EBS method detected additional genes related to colon cancer that were missed by other methods.This novel empirical Bayes procedure is advantageous over our earlier proposed empirical Bayes adjustments due to the following reasons: (i) it offers an automatic screening of the p-values the user may obtain from a univariate (i.e., gene by gene) analysis package making it extremely easy to use for a non-statistician, (ii) since it applies to the p-values, the tests do not have to be t-tests; in particular they could be F-tests which might arise in certain ANOVA formulations with expression data or even nonparametric tests, (iii) the empirical Bayes adjustment uses nonparametric function estimation techniques to estimate the marginal density of the transformed p-values rather than using a parametric model for the prior distribution and is therefore robust against model mis-specification. AVAILABILITY: R code for EBS is available from the authors upon request. SUPPLEMENTARY INFORMATION: http://www.stat.uga.edu/~datta/EBS/supp.htm     

Feature Selection for Gene Expression Using Model-Based Entropy

Gene expression data usually contain a large number of genes but a small number of samples. Feature selection for gene expression data aims at finding a set of genes that best discriminate biological samples of different types. Using machine learning techniques, traditional gene selection based on empirical mutual information suffers the data sparseness issue due to the small number of samples. To overcome the sparseness issue, we propose a model-based approach to estimate the entropy of class variables on the model, instead of on the data themselves. Here, we use multivariate normal distributions to fit the data, because multivariate normal distributions have maximum entropy among all real-valued distributions with a specified mean and standard deviation and are widely used to approximate various distributions. Given that the data follow a multivariate normal distribution, since the conditional distribution of class variables given the selected features is a normal distribution, its entropy can be computed with the log-determinant of its covariance matrix. Because of the large number of genes, the computation of all possible log-determinants is not efficient. We propose several algorithms to largely reduce the computational cost. The experiments on seven gene data sets and the comparison with other five approaches show the accuracy of the multivariate Gaussian generative model for feature selection, and the efficiency of our algorithms.     

An empirical bayes adjustment to increase the sensitivity of detecting differentially expressed genes in microarray experiments

MOTIVATION: Detection of differentially expressed genes is one of the major goals of microarray experiments. Pairwise comparison for each gene is not appropriate without controlling the overall (experimentwise) type 1 error rate. Dudoit et al. have advocated use of permutation-based step-down P-value adjustments to correct the observed significance levels for the individual (i.e. for each gene) two sample t-tests. RESULTS: In this paper, we consider an ANOVA formulation of the gene expression levels corresponding to multiple tissue types. We provide resampling-based step-down adjustments to correct the observed significance levels for the individual ANOVA t-tests for each gene and for each pair of tissue type comparisons. More importantly, we introduce a novel empirical Bayes adjustment to the t-test statistics that can be incorporated into the step-down procedure. Using simulated data, we show that the empirical Bayes adjustment improved the sensitivity of detecting differentially expressed genes up to 16%, while maintaining a high level of specificity. This adjustment also reduces the false non-discovery rate to some degree at the cost of a modest increase in the false discovery rate. We illustrate our approach using a human colon cancer dataset consisting of oligonucleotide arrays of normal, adenoma and carcinoma cells. The number of genes with differential expression level declared statistically significant was about 50 when comparing normal to adenoma cells and about five when comparing adenoma to carcinoma cells. This list includes genes previously known to be associated with colon cancer as well as some novel genes. AVAILABILITY: R code for the empirical Bayes adjustment and step-down P-value calculation via resampling are available from the supplementary web-site. Supplementary information: http://www.mathstat.gsu.edu/~matsnd/EB/supp.htm     

ESPD: a pattern detection model underlying gene expression profiles

MOTIVATION: DNA arrays permit rapid, large-scale screening for patterns of gene expression and simultaneously yield the expression levels of thousands of genes for samples. The number of samples is usually limited, and such datasets are very sparse in high-dimensional gene space. Furthermore, most of the genes collected may not necessarily be of interest and uncertainty about which genes are relevant makes it difficult to construct an informative gene space. Unsupervised empirical sample pattern discovery and informative genes identification of such sparse high-dimensional datasets present interesting but challenging problems. RESULTS: A new model called empirical sample pattern detection (ESPD) is proposed to delineate pattern quality with informative genes. By integrating statistical metrics, data mining and machine learning techniques, this model dynamically measures and manipulates the relationship between samples and genes while conducting an iterative detection of informative space and the empirical pattern. The performance of the proposed method with various array datasets is illustrated.     

Bayesian model averaging: development of an improved multi-class, gene selection and classification tool for microarray data

MOTIVATION: Selecting a small number of relevant genes for accurate classification of samples is essential for the development of diagnostic tests. We present the Bayesian model averaging (BMA) method for gene selection and classification of microarray data. Typical gene selection and classification procedures ignore model uncertainty and use a single set of relevant genes (model) to predict the class. BMA accounts for the uncertainty about the best set to choose by averaging over multiple models (sets of potentially overlapping relevant genes). RESULTS: We have shown that BMA selects smaller numbers of relevant genes (compared with other methods) and achieves a high prediction accuracy on three microarray datasets. Our BMA algorithm is applicable to microarray datasets with any number of classes, and outputs posterior probabilities for the selected genes and models. Our selected models typically consist of only a few genes. The combination of high accuracy, small numbers of genes and posterior probabilities for the predictions should make BMA a powerful tool for developing diagnostics from expression data. AVAILABILITY: The source codes and datasets used are available from our Supplementary website.     

A Bayesian network classification methodology for gene expression data.

We present new techniques for the application of a Bayesian network learning framework to the problem of classifying gene expression data. The focus on classification permits us to develop techniques that address in several ways the complexities of learning Bayesian nets. Our classification model reduces the Bayesian network learning problem to the problem of learning multiple subnetworks, each consisting of a class label node and its set of parent genes. We argue that this classification model is more appropriate for the gene expression domain than are other structurally similar Bayesian network classification models, such as Naive Bayes and Tree Augmented Naive Bayes (TAN), because our model is consistent with prior domain experience suggesting that a relatively small number of genes, taken in different combinations, is required to predict most clinical classes of interest. Within this framework, we consider two different approaches to identifying parent sets which are supported by the gene expression observations and any other currently available evidence. One approach employs a simple greedy algorithm to search the universe of all genes; the second approach develops and applies a gene selection algorithm whose results are incorporated as a prior to enable an exhaustive search for parent sets over a restricted universe of genes. Two other significant contributions are the construction of classifiers from multiple, competing Bayesian network hypotheses and algorithmic methods for normalizing and binning gene expression data in the absence of prior expert knowledge. Our classifiers are developed under a cross validation regimen and then validated on corresponding out-of-sample test sets. The classifiers attain a classification rate in excess of 90% on out-of-sample test sets for two publicly available datasets. We present an extensive compilation of results reported in the literature for other classification methods run against these same two datasets. Our results are comparable to, or better than, any we have found reported for these two sets, when a train-test protocol as stringent as ours is followed.     

A comparative study of feature selection and multiclass classification methods for tissue classification based on gene expression

This paper studies the problem of building multiclass classifiers for tissue classification based on gene expression. The recent development of microarray technologies has enabled biologists to quantify gene expression of tens of thousands of genes in a single experiment. Biologists have begun collecting gene expression for a large number of samples. One of the urgent issues in the use of microarray data is to develop methods for characterizing samples based on their gene expression. The most basic step in the research direction is binary sample classification, which has been studied extensively over the past few years. This paper investigates the next step-multiclass classification of samples based on gene expression. The characteristics of expression data (e.g. large number of genes with small sample size) makes the classification problem more challenging. The process of building multiclass classifiers is divided into two components: (i) selection of the features (i.e. genes) to be used for training and testing and (ii) selection of the classification method. This paper compares various feature selection methods as well as various state-of-the-art classification methods on various multiclass gene expression datasets. Our study indicates that multiclass classification problem is much more difficult than the binary one for the gene expression datasets. The difficulty lies in the fact that the data are of high dimensionality and that the sample size is small. The classification accuracy appears to degrade very rapidly as the number of classes increases. In particular, the accuracy was very low regardless of the choices of the methods for large-class datasets (e.g. NCI60 and GCM). While increasing the number of samples is a plausible solution to the problem of accuracy degradation, it is important to develop algorithms that are able to analyze effectively multiple-class expression data for these special datasets.     

Borrowing information across genes and experiments for improved error variance estimation in microarray data analysis

Statistical inference for microarray experiments usually involves the estimation of error variance for each gene. Because the sample size available for each gene is often low, the usual unbiased estimator of the error variance can be unreliable. Shrinkage methods, including empirical Bayes approaches that borrow information across genes to produce more stable estimates, have been developed in recent years. Because the same microarray platform is often used for at least several experiments to study similar biological systems, there is an opportunity to improve variance estimation further by borrowing information not only across genes but also across experiments. We propose a lognormal model for error variances that involves random gene effects and random experiment effects. Based on the model, we develop an empirical Bayes estimator of the error variance for each combination of gene and experiment and call this estimator BAGE because information is Borrowed Across Genes and Experiments. A permutation strategy is used to make inference about the differential expression status of each gene. Simulation studies with data generated from different probability models and real microarray data show that our method outperforms existing approaches.     

An empirical Bayes approach to inferring large-scale gene association networks

MOTIVATION: Genetic networks are often described statistically using graphical models (e.g. Bayesian networks). However, inferring the network structure offers a serious challenge in microarray analysis where the sample size is small compared to the number of considered genes. This renders many standard algorithms for graphical models inapplicable, and inferring genetic networks an 'ill-posed' inverse problem. METHODS: We introduce a novel framework for small-sample inference of graphical models from gene expression data. Specifically, we focus on the so-called graphical Gaussian models (GGMs) that are now frequently used to describe gene association networks and to detect conditionally dependent genes. Our new approach is based on (1) improved (regularized) small-sample point estimates of partial correlation, (2) an exact test of edge inclusion with adaptive estimation of the degree of freedom and (3) a heuristic network search based on false discovery rate multiple testing. Steps (2) and (3) correspond to an empirical Bayes estimate of the network topology. RESULTS: Using computer simulations, we investigate the sensitivity (power) and specificity (true negative rate) of the proposed framework to estimate GGMs from microarray data. This shows that it is possible to recover the true network topology with high accuracy even for small-sample datasets. Subsequently, we analyze gene expression data from a breast cancer tumor study and illustrate our approach by inferring a corresponding large-scale gene association network for 3883 genes.     

Differential expression analysis of multifactor RNA-Seq experiments with respect to biological variation

A flexible statistical framework is developed for the analysis of read counts from RNA-Seq gene expression studies. It provides the ability to analyse complex experiments involving multiple treatment conditions and blocking variables while still taking full account of biological variation. Biological variation between RNA samples is estimated separately from the technical variation associated with sequencing technologies. Novel empirical Bayes methods allow each gene to have its own specific variability, even when there are relatively few biological replicates from which to estimate such variability. The pipeline is implemented in the edgeR package of the Bioconductor project. A case study analysis of carcinoma data demonstrates the ability of generalized linear model methods (GLMs) to detect differential expression in a paired design, and even to detect tumour-specific expression changes. The case study demonstrates the need to allow for gene-specific variability, rather than assuming a common dispersion across genes or a fixed relationship between abundance and variability. Genewise dispersions de-prioritize genes with inconsistent results and allow the main analysis to focus on changes that are consistent between biological replicates. Parallel computational approaches are developed to make non-linear model fitting faster and more reliable, making the application of GLMs to genomic data more convenient and practical. Simulations demonstrate the ability of adjusted profile likelihood estimators to return accurate estimators of biological variability in complex situations. When variation is gene-specific, empirical Bayes estimators provide an advantageous compromise between the extremes of assuming common dispersion or separate genewise dispersion. The methods developed here can also be applied to count data arising from DNA-Seq applications, including ChIP-Seq for epigenetic marks and DNA methylation analyses.     

Estimating effect sizes of differentially expressed genes for power and sample-size assessments in microarray experiments

Summary In microarray screening for differentially expressed genes using multiple testing, assessment of power or sample size is of particular importance to ensure that few relevant genes are removed from further consideration prematurely. In this assessment, adequate estimation of the effect sizes of differentially expressed genes is crucial because of its substantial impact on power and sample‐size estimates. However, conventional methods using top genes with largest observed effect sizes would be subject to overestimation due to random variation. In this article, we propose a simple estimation method based on hierarchical mixture models with a nonparametric prior distribution to accommodate random variation and possible large diversity of effect sizes across differential genes, separated from nuisance, nondifferential genes. Based on empirical Bayes estimates of effect sizes, the power and false discovery rate (FDR) can be estimated to monitor them simultaneously in gene screening. We also propose a power index that concerns selection of top genes with largest effect sizes, called partial power. This new power index could provide a practical compromise for the difficulty in achieving high levels of usual overall power as confronted in many microarray experiments. Applications to two real datasets from cancer clinical studies are provided.     

Empirical Bayes estimation of gene-specific effects in micro-array research

Micro-array technology allows investigators the opportunity to measure expression levels of thousands of genes simultaneously. However, investigators are also faced with the challenge of simultaneous estimation of gene expression differences for thousands of genes with very small sample sizes. Traditional estimators of differences between treatment means (ordinary least squares estimators or OLS) are not the best estimators if interest is in estimation of gene expression differences for an ensemble of genes. In the case that gene expression differences are regarded as exchangeable samples from a common population, estimators are available that result in much smaller average mean-square error across the population of gene expression difference estimates. We have simulated the application of such an estimator, namely an empirical Bayes (EB) estimator of random effects in a hierarchical linear model (normal-normal). Simulation results revealed mean-square error as low as 0.05 times the mean-square error of OLS estimators (i.e., the difference between treatment means). We applied the analysis to an example dataset as a demonstration of the shrinkage of EB estimators and of the reduction in mean-square error, i.e., increase in precision, associated with EB estimators in this analysis. The method described here is available in software that is available at .     

A decomposition model to track gene expression signatures: preview on observer-independent classification of ovarian cancer

MOTIVATION: A number of algorithms and analytical models have been employed to reduce the multidimensional complexity of DNA array data and attempt to extract some meaningful interpretation of the results. These include clustering, principal components analysis, self-organizing maps, and support vector machine analysis. Each method assumes an implicit model for the data, many of which separate genes into distinct clusters defined by similar expression profiles in the samples tested. A point of concern is that many genes may be involved in a number of distinct behaviours, and should therefore be modelled to fit into as many separate clusters as detected in the multidimensional gene expression space. The analysis of gene expression data using a decomposition model that is independent of the observer involved would be highly beneficial to improve standard and reproducible classification of clinical and research samples. RESULTS: We present a variational independent component analysis (ICA) method for reducing high dimensional DNA array data to a smaller set of latent variables, each associated with a gene signature. We present the results of applying the method to data from an ovarian cancer study, revealing a number of tissue type-specific and tissue type-independent gene signatures present in varying amounts among the samples surveyed. The observer independent results of such molecular analysis of biological samples could help identify patients who would benefit from different treatment strategies. We further explore the application of the model to similar high-throughput studies.     

Differential gene expression detection and sample classification using penalized linear regression models

Differential gene expression detection and sample classification using microarray data have received much research interest recently. Owing to the large number of genes p and small number of samples n (p > n), microarray data analysis poses big challenges for statistical analysis. An obvious problem owing to the 'large p small n' is over-fitting. Just by chance, we are likely to find some non-differentially expressed genes that can classify the samples very well. The idea of shrinkage is to regularize the model parameters to reduce the effects of noise and produce reliable inferences. Shrinkage has been successfully applied in the microarray data analysis. The SAM statistics proposed by Tusher et al. and the 'nearest shrunken centroid' proposed by Tibshirani et al. are ad hoc shrinkage methods. Both methods are simple, intuitive and prove to be useful in empirical studies. Recently Wu proposed the penalized t/F-statistics with shrinkage by formally using the (1) penalized linear regression models for two-class microarray data, showing good performance. In this paper we systematically discussed the use of penalized regression models for analyzing microarray data. We generalize the two-class penalized t/F-statistics proposed by Wu to multi-class microarray data. We formally derive the ad hoc shrunken centroid used by Tibshirani et al. using the (1) penalized regression models. And we show that the penalized linear regression models provide a rigorous and unified statistical framework for sample classification and differential gene expression detection.     

Reliable gene signatures for microarray classification: assessment of stability and performance

MOTIVATION: Two important questions for the analysis of gene expression measurements from different sample classes are (1) how to classify samples and (2) how to identify meaningful gene signatures (ranked gene lists) exhibiting the differences between classes and sample subsets. Solutions to both questions have immediate biological and biomedical applications. To achieve optimal classification performance, a suitable combination of classifier and gene selection method needs to be specifically selected for a given dataset. The selected gene signatures can be unstable and the resulting classification accuracy unreliable, particularly when considering different subsets of samples. Both unstable gene signatures and overestimated classification accuracy can impair biological conclusions. METHODS: We address these two issues by repeatedly evaluating the classification performance of all models, i.e. pairwise combinations of various gene selection and classification methods, for random subsets of arrays (sampling). A model score is used to select the most appropriate model for the given dataset. Consensus gene signatures are constructed by extracting those genes frequently selected over many samplings. Sampling additionally permits measurement of the stability of the classification performance for each model, which serves as a measure of model reliability. RESULTS: We analyzed a large gene expression dataset with 78 measurements of four different cartilage sample classes. Classifiers trained on subsets of measurements frequently produce models with highly variable performance. Our approach provides reliable classification performance estimates via sampling. In addition to reliable classification performance, we determined stable consensus signatures (i.e. gene lists) for sample classes. Manual literature screening showed that these genes are highly relevant to our gene expression experiment with osteoarthritic cartilage. We compared our approach to others based on a publicly available dataset on breast cancer. AVAILABILITY: R package at http://www.bio.ifi.lmu.de/~davis/edaprakt     

Intensity-based hierarchical Bayes method improves testing for differentially expressed genes in microarray experiments

Background  

The small sample sizes often used for microarray experiments result in poor estimates of variance if each gene is considered independently. Yet accurately estimating variability of gene expression measurements in microarray experiments is essential for correctly identifying differentially expressed genes. Several recently developed methods for testing differential expression of genes utilize hierarchical Bayesian models to "pool" information from multiple genes. We have developed a statistical testing procedure that further improves upon current methods by incorporating the well-documented relationship between the absolute gene expression level and the variance of gene expression measurements into the general empirical Bayes framework.