Microarrays: handling the deluge of data and extracting reliable information

Application of powerful, high-throughput genomics technologies is becoming more common and these technologies are evolving at a rapid pace. Genomics facilities are being established in major research institutions to produce inexpensive, customized cDNA microarrays that are accessible to researchers in a broad range of fields. These high-throughput platforms have generated a massive onslaught of data, which threatens to overwhelm researchers. Although microarrays show great promise, the technology has not matured to the point of consistently generating robust and reliable data when used in the average laboratory. This article addresses several aspects related to the handling of the deluge of microarray data and extracting reliable information from these data. We review the essential elements of data acquisition, data processing and data analysis, and briefly discuss issues related to the quality, validation and storage of data. Our goal is to point out some of the problems that must be overcome before this promising technology can achieve its full potential.     

Storage and retrieval of microarray data and open source microarray database software

Microarray technology has been widely adopted by researchers who use both home-made microarrays and microarrays purchased from commercial vendors. Associated with the adoption of this technology has been a deluge of complex data, both from the microarrays themselves, and also in the form of associated meta data, such as gene annotation information, the properties and treatment of biological samples, and the data transformation and analysis steps taken downstream. In addition, standards for annotation and data exchange have been proposed, and are now being adopted by journals and funding agencies alike. The coupling of large quantities of complex data with extensive and complex standards require all but the most small-scale of microarray users to have access to a robust and scaleable database with various tools. In this review, we discuss some of the desirable properties of such a database, and look at the features of several freely available alternatives.     

Microarray data quality analysis: lessons from the AFGC project

Genome-wide expression profiling with DNA microarrays has and will provide a great deal of data to the plant scientific community. However, reliability concerns have required the development data quality tests for common systematic biases. Fortunately, most large-scale systematic biases are detectable and some are correctable by normalization. Technical replication experiments and statistical surveys indicate that these biases vary widely in severity and appearance. As a result, no single normalization or correction method currently available is able to address all the issues. However, careful sequence selection, array design, experimental design and experimental annotation can substantially improve the quality and biological of microarray data. In this review, we discuss these issues with reference to examples from the Arabidopsis Functional Genomics Consortium (AFGC) microarray project.     

Microarray data quality analysis: lessons from the AFGC project. Arabidopsis Functional Genomics Consortium

Genome-wide expression profiling with DNA microarrays has and will provide a great deal of data to the plant scientific community. However, reliability concerns have required the development data quality tests for common systematic biases. Fortunately, most large-scale systematic biases are detectable and some are correctable by normalization. Technical replication experiments and statistical surveys indicate that these biases vary widely in severity and appearance. As a result, no single normalization or correction method currently available is able to address all the issues. However, careful sequence selection, array design, experimental design and experimental annotation can substantially improve the quality and biological of microarray data. In this review, we discuss these issues with reference to examples from the Arabidopsis Functional Genomics Consortium (AFGC) microarray project.     

New normalization methods for cDNA microarray data

MOTIVATION: The focus of this paper is on two new normalization methods for cDNA microarrays. After the image analysis has been performed on a microarray and before differentially expressed genes can be detected, some form of normalization must be applied to the microarrays. Normalization removes biases towards one or other of the fluorescent dyes used to label each mRNA sample allowing for proper evaluation of differential gene expression. RESULTS: The two normalization methods that we present here build on previously described non-linear normalization techniques. We extend these techniques by firstly introducing a normalization method that deals with smooth spatial trends in intensity across microarrays, an important issue that must be dealt with. Secondly we deal with normalization of a new type of cDNA microarray experiment that is coming into prevalence, the small scale specialty or 'boutique' array, where large proportions of the genes on the microarrays are expected to be highly differentially expressed. AVAILABILITY: The normalization methods described in this paper are available via http://www.pi.csiro.au/gena/ in a software suite called tRMA: tools for R Microarray Analysis upon request of the authors. Images and data used in this paper are also available via the same link.     

A practical false discovery rate approach to identifying patterns of differential expression in microarray data

SUMMARY: Searching for differentially expressed genes is one of the most common applications for microarrays, yet statistically there are difficult hurdles to achieving adequate rigor and practicality. False discovery rate (FDR) approaches have become relatively standard; however, how to define and control the FDR has been hotly debated. Permutation estimation approaches such as SAM and PaGE can be effective; however, they leave much room for improvement. We pursue the permutation estimation method and describe a convenient definition for the FDR that can be estimated in a straightforward manner. We then discuss issues regarding the choice of statistic and data transformation. It is impossible to optimize the power of any statistic for thousands of genes simultaneously, and we look at the practical consequences of this. For example, the log transform can both help and hurt at the same time, depending on the gene. We examine issues surrounding the SAM 'fudge factor' parameter, and how to handle these issues by optimizing with respect to power.     

Analyzing time series gene expression data

MOTIVATION: Time series expression experiments are an increasingly popular method for studying a wide range of biological systems. However, when analyzing these experiments researchers face many new computational challenges. Algorithms that are specifically designed for time series experiments are required so that we can take advantage of their unique features (such as the ability to infer causality from the temporal response pattern) and address the unique problems they raise (e.g. handling the different non-uniform sampling rates). RESULTS: We present a comprehensive review of the current research in time series expression data analysis. We divide the computational challenges into four analysis levels: experimental design, data analysis, pattern recognition and networks. For each of these levels, we discuss computational and biological problems at that level and point out some of the methods that have been proposed to deal with these issues. Many open problems in all these levels are discussed. This review is intended to serve as both, a point of reference for experimental biologists looking for practical solutions for analyzing their data, and a starting point for computer scientists interested in working on the computational problems related to time series expression analysis.     

Statistical design and the analysis of gene expression microarray data

Gene expression microarrays are an innovative technology with enormous promise to help geneticists explore and understand the genome. Although the potential of this technology has been clearly demonstrated, many important and interesting statistical questions persist. We relate certain features of microarrays to other kinds of experimental data and argue that classical statistical techniques are appropriate and useful. We advocate greater attention to experimental design issues and a more prominent role for the ideas of statistical inference in microarray studies.     

Expression Profiling with Oligonucleotide Arrays: Technologies and Applications for Neurobiology

DNA microarrays have been used in applications ranging from the assignment of gene function to analytical uses in prognostics. However, the detection sensitivity, cross hybridization, and reproducibility of these arrays can affect experimental design and data interpretation. Moreover, several technologies are available for fabrication of oligonucleotide microarrays. We review these technologies and performance attributes and, with data sets generated from human brain RNA, present statistical tools and methods to analyze data quality and to mine and visualize the data. Our data show high reproducibility and should allow an investigator to discern biological and regional variability from differential expression. Although we have used brain RNA as a model system to illustrate some of these points, the oligonucleotide arrays and methods employed in this study can be used with cell lines, tissue sections, blood, and other fluids. To further demonstrate this point, we provide data generated from total RNA sample sizes of 200 ng.     

Beyond hairballs: The use of quantitative mass spectrometry data to understand protein-protein interactions

The past 10years have witnessed a dramatic proliferation in the availability of protein interaction data. However, for interaction mapping based on affinity purification coupled with mass spectrometry (AP-MS), there is a wealth of information present in the datasets that often goes unrecorded in public repositories, and as such remains largely unexplored. Further, how this type of data is represented and used by bioinformaticians has not been well established. Here, we point out some common mistakes in how AP-MS data are handled, and describe how protein complex organization and interaction dynamics can be inferred using quantitative AP-MS approaches.     

Spot shape modelling and data transformations for microarrays

MOTIVATION: To study lowly expressed genes in microarray experiments, it is useful to increase the photometric gain in the scanning. However, a large gain may cause some pixels for highly expressed genes to become saturated. Spatial statistical models that model spot shapes on the pixel level may be used to infer information about the saturated pixel intensities. Other possible applications for spot shape models include data quality control and accurate determination of spot centres and spot diameters. RESULTS: Spatial statistical models for spotted microarrays are studied including pixel level transformations and spot shape models. The models are applied to a dataset from 50mer oligonucleotide microarrays with 452 selected Arabidopsis genes. Logarithmic, Box-Cox and inverse hyperbolic sine transformations are compared in combination with four spot shape models: a cylindric plateau shape, an isotropic Gaussian distribution and a difference of two-scaled Gaussian distribution suggested in the literature, as well as a proposed new polynomial-hyperbolic spot shape model. A substantial improvement is obtained for the dataset studied by the polynomial-hyperbolic spot shape model in combination with the Box-Cox transformation. The spatial statistical models are used to correct spot measurements with saturation by extrapolating the censored data. AVAILABILITY: Source code for R is available at http://www.matfys.kvl.dk/~ekstrom/spotshapes/     

FLAME,a novel fuzzy clustering method for the analysis of DNA microarray data

Background  

Data clustering analysis has been extensively applied to extract information from gene expression profiles obtained with DNA microarrays. To this aim, existing clustering approaches, mainly developed in computer science, have been adapted to microarray data analysis. However, previous studies revealed that microarray datasets have very diverse structures, some of which may not be correctly captured by current clustering methods. We therefore approached the problem from a new starting point, and developed a clustering algorithm designed to capture dataset-specific structures at the beginning of the process.     

Approximate variance-stabilizing transformations for gene-expression microarray data

MOTIVATION: A variance stabilizing transformation for microarray data was recently introduced independently by several research groups. This transformation has sometimes been called the generalized logarithm or glog transformation. In this paper, we derive several alternative approximate variance stabilizing transformations that may be easier to use in some applications. RESULTS: We demonstrate that the started-log and the log-linear-hybrid transformation families can produce approximate variance stabilizing transformations for microarray data that are nearly as good as the generalized logarithm (glog) transformation. These transformations may be more convenient in some applications.     

Design and analysis of experiments with high throughput biological assay data

The design and analysis of experiments using gene expression microarrays is a topic of considerable current research, and work is beginning to appear on the analysis of proteomics and metabolomics data by mass spectrometry and NMR spectroscopy. The literature in this area is evolving rapidly, and commercial software for analysis of array or proteomics data is rarely up to date, and is essentially nonexistent for metabolomics data. In this paper, I review some of the issues that should concern any biologists planning to use such high-throughput biological assay data in an experimental investigation. Technical details are kept to a minimum, and may be found in the referenced literature, as well as in the many excellent papers which space limitations prevent my describing. There are usually a number of viable options for design and analysis of such experiments, but unfortunately, there are even more non-viable ones that have been used even in the published literature. This is an area in which up-to-date knowledge of the literature is indispensable for efficient and effective design and analysis of these experiments. In general, we concentrate on relatively simple analyses, often focusing on identifying differentially expressed genes and the comparable issues in mass spectrometry and NMR spectroscopy (consistent differences in peak heights or areas for example). Complex multivariate and pattern recognition methods also need much attention, but the issues we describe in this paper must be dealt with first. The literature on analysis of proteomics and metabolomics data is as yet sparse, so the main focus of this paper will be on methods devised for analysis of gene expression data that generalize to proteomics and metabolomics, with some specific comments near the end on analysis of metabolomics data by mass spectrometry and NMR spectroscopy.     

The high-throughput yeast deletion fitness data and the theories of dominance

The development of high-throughput fitness measurement methods provides unprecedented power to test evolutionary theories. However, with this comes new challenges regarding data quality and data analysis. We illustrate this by reanalysing the fitness distribution in several environments of yeast mutants (homo- and heterozygous) from the yeast deletion project. Originally created to study functional properties of genes, evolutionary biologists took advantage of this database to study evolutionary questions, such as dominance for fitness of mutations. We uncover several problems in this data set strongly affecting these questions that have remained unnoticed despite the numerous studies based on it. High-throughput methodologies are necessarily challenging, both experimentally and for data analysis: our point is not to criticize these approaches, but to pinpoint these challenges and to propose several improvements that may help avoid several shortcomings. Further, in the light of this finding, we question the conclusions regarding theories of dominance that have been made using this data set. We show that the data on deletion of small effects are not sufficiently reliable to be informative on this question. On the other hand, deletions of large effect exhibit no correlation between homo- and heterozygous fitness effects, a pattern that sheds new light on the h-s correlation issue, with several consequences for the debate over the different theories of dominance.     

BAMarray™: Java software for Bayesian analysis of variance for microarray data

Background  

DNA microarrays open up a new horizon for studying the genetic determinants of disease. The high throughput nature of these arrays creates an enormous wealth of information, but also poses a challenge to data analysis. Inferential problems become even more pronounced as experimental designs used to collect data become more complex. An important example is multigroup data collected over different experimental groups, such as data collected from distinct stages of a disease process. We have developed a method specifically addressing these issues termed Bayesian ANOVA for microarrays (BAM). The BAM approach uses a special inferential regularization known as spike-and-slab shrinkage that provides an optimal balance between total false detections and total false non-detections. This translates into more reproducible differential calls. Spike and slab shrinkage is a form of regularization achieved by using information across all genes and groups simultaneously.     

Model-based clustering and data transformations for gene expression data.

MOTIVATION: Clustering is a useful exploratory technique for the analysis of gene expression data. Many different heuristic clustering algorithms have been proposed in this context. Clustering algorithms based on probability models offer a principled alternative to heuristic algorithms. In particular, model-based clustering assumes that the data is generated by a finite mixture of underlying probability distributions such as multivariate normal distributions. The issues of selecting a 'good' clustering method and determining the 'correct' number of clusters are reduced to model selection problems in the probability framework. Gaussian mixture models have been shown to be a powerful tool for clustering in many applications. RESULTS: We benchmarked the performance of model-based clustering on several synthetic and real gene expression data sets for which external evaluation criteria were available. The model-based approach has superior performance on our synthetic data sets, consistently selecting the correct model and the number of clusters. On real expression data, the model-based approach produced clusters of quality comparable to a leading heuristic clustering algorithm, but with the key advantage of suggesting the number of clusters and an appropriate model. We also explored the validity of the Gaussian mixture assumption on different transformations of real data. We also assessed the degree to which these real gene expression data sets fit multivariate Gaussian distributions both before and after subjecting them to commonly used data transformations. Suitably chosen transformations seem to result in reasonable fits. AVAILABILITY: MCLUST is available at http://www.stat.washington.edu/fraley/mclust. The software for the diagonal model is under development. CONTACT: kayee@cs.washington.edu. SUPPLEMENTARY INFORMATION: http://www.cs.washington.edu/homes/kayee/model.