Cross-platform microarray data normalisation for regulatory network inference

Background

Inferring Gene Regulatory Networks (GRNs) from time course microarray data suffers from the dimensionality problem created by the short length of available time series compared to the large number of genes in the network. To overcome this, data integration from diverse sources is mandatory. Microarray data from different sources and platforms are publicly available, but integration is not straightforward, due to platform and experimental differences.

Methods

We analyse here different normalisation approaches for microarray data integration, in the context of reverse engineering of GRN quantitative models. We introduce two preprocessing approaches based on existing normalisation techniques and provide a comprehensive comparison of normalised datasets.

Conclusions

Results identify a method based on a combination of Loess normalisation and iterative K-means as best for time series normalisation for this problem.     

A comparison on effects of normalisations in the detection of differentially expressed genes

Background  

Various normalisation techniques have been developed in the context of microarray analysis to try to correct expression measurements for experimental bias and random fluctuations. Major techniques include: total intensity normalisation; intensity dependent normalisation; and variance stabilising normalisation. The aim of this paper is to discuss the impact of normalisation techniques for two-channel array technology on the process of identification of differentially expressed genes.     

Analysis of a simulated microarray dataset: Comparison of methods for data normalisation and detection of differential expression (Open Access publication)

Microarrays allow researchers to measure the expression of thousands of genes in a single experiment. Before statistical comparisons can be made, the data must be assessed for quality and normalisation procedures must be applied, of which many have been proposed. Methods of comparing the normalised data are also abundant, and no clear consensus has yet been reached. The purpose of this paper was to compare those methods used by the EADGENE network on a very noisy simulated data set. With the a priori knowledge of which genes are differentially expressed, it is possible to compare the success of each approach quantitatively. Use of an intensity-dependent normalisation procedure was common, as was correction for multiple testing. Most variety in performance resulted from differing approaches to data quality and the use of different statistical tests. Very few of the methods used any kind of background correction. A number of approaches achieved a success rate of 95% or above, with relatively small numbers of false positives and negatives. Applying stringent spot selection criteria and elimination of data did not improve the false positive rate and greatly increased the false negative rate. However, most approaches performed well, and it is encouraging that widely available techniques can achieve such good results on a very noisy data set.     

A model-based analysis of microarray experimental error and normalisation

A statistical model is proposed for the analysis of errors in microarray experiments and is employed in the analysis and development of a combined normalisation regime. Through analysis of the model and two-dye microarray data sets, this study found the following. The systematic error introduced by microarray experiments mainly involves spot intensity-dependent, feature-specific and spot position-dependent contributions. It is difficult to remove all these errors effectively without a suitable combined normalisation operation. Adaptive normalisation using a suitable regression technique is more effective in removing spot intensity-related dye bias than self-normalisation, while regional normalisation (block normalisation) is an effective way to correct spot position-dependent errors. However, dye-flip replicates are necessary to remove feature-specific errors, and also allow the analyst to identify the experimentally introduced dye bias contained in non-self-self data sets. In this case, the bias present in the data sets may include both experimentally introduced dye bias and the biological difference between two samples. Self-normalisation is capable of removing dye bias without identifying the nature of that bias. The performance of adaptive normalisation, on the other hand, depends on its ability to correctly identify the dye bias. If adaptive normalisation is combined with an effective dye bias identification method then there is no systematic difference between the outcomes of the two methods.     

Application of z-score transformation to Affymetrix data

Z-score transformation has been successfully used as a normalisation procedure for microarray data generated using radioactively labelled probes with spotted cDNA arrays. One of the advantages of the z-score transformation method is that it provides a way of standardising data across a wide range of experiments and allows the comparison of microarray data independent of the original hybridisation intensities. The feasibility of applying z-score transformation to other types of linear microarray data, specifically that generated using fluorescently labelled probes with Affymetrix chips, was tested in three separate scenarios and is discussed here. In the first scenario, Affymetrix data from the NCBI (National Center for Biotechnology Information) GEO (Gene Expression Omnibus) database was used to demonstrate that z-score transformation preserved the essential phylogenetic grouping between primate species' fibroblast gene expression baseline measurements. The second scenario employed z-score transformation on data consisting of a series of genes spiked-in at known concentrations and arrayed in a Latin square format. We were able to reconstruct the entire set of spike-in concentration curves without prior knowledge of their format by using z-score transformation as the normalisation process. Finally, we show that z-score transformed data maintains the integrity of separate samples from different experiments and laboratories, as demonstrated by accurate grouping of clustered data according to sample identity. We conclude that data normalised by z-score transformation can be easily used with Affymetrix data without noticeable loss of information content. Z-score transformation provides a useful tool for comparisons between experiments and between laboratories that use the Affymetrix platform.     

The EADGENE Microarray Data Analysis Workshop (Open Access publication)

Microarray analyses have become an important tool in animal genomics. While their use is becoming widespread, there is still a lot of ongoing research regarding the analysis of microarray data. In the context of a European Network of Excellence, 31 researchers representing 14 research groups from 10 countries performed and discussed the statistical analyses of real and simulated 2-colour microarray data that were distributed among participants. The real data consisted of 48 microarrays from a disease challenge experiment in dairy cattle, while the simulated data consisted of 10 microarrays from a direct comparison of two treatments (dye-balanced). While there was broader agreement with regards to methods of microarray normalisation and significance testing, there were major differences with regards to quality control. The quality control approaches varied from none, through using statistical weights, to omitting a large number of spots or omitting entire slides. Surprisingly, these very different approaches gave quite similar results when applied to the simulated data, although not all participating groups analysed both real and simulated data. The workshop was very successful in facilitating interaction between scientists with a diverse background but a common interest in microarray analyses.     

A comparison of normalization techniques for microRNA microarray data

Normalization of expression levels applied to microarray data can help in reducing measurement error. Different methods, including cyclic loess, quantile normalization and median or mean normalization, have been utilized to normalize microarray data. Although there is considerable literature regarding normalization techniques for mRNA microarray data, there are no publications comparing normalization techniques for microRNA (miRNA) microarray data, which are subject to similar sources of measurement error. In this paper, we compare the performance of cyclic loess, quantile normalization, median normalization and no normalization for a single-color microRNA microarray dataset. We show that the quantile normalization method works best in reducing differences in miRNA expression values for replicate tissue samples. By showing that the total mean squared error are lowest across almost all 36 investigated tissue samples, we are assured that the bias correction provided by quantile normalization is not outweighed by additional error variance that can arise from a more complex normalization method. Furthermore, we show that quantile normalization does not achieve these results by compression of scale.     

Replication, Variation and Normalisation in Microarray Experiments

INTRODUCTION: Microarray experiments often have complex designs that include sample pooling, biological and technical replication, sample pairing and dye-swapping. This article demonstrates how statistical modelling can illuminate issues in the design and analysis of microarray experiments, and this information can then be used to plan effective studies. METHODS: A very detailed statistical model for microarray data is introduced, to show the possible sources of variation that are present in even the simplest microarray experiments. Based on this model, the efficacy of common experimental designs, normalisation methodologies and analyses is determined. RESULTS: When the cost of the arrays is high compared with the cost of samples, sample pooling and spot replication are shown to be efficient variance reduction methods, whereas technical replication of whole arrays is demonstrated to be very inefficient. Dye-swap designs can use biological replicates rather than technical replicates to improve efficiency and simplify analysis. When the cost of samples is high and technical variation is a major portion of the error, technical replication can be cost effective. Normalisation by centreing on a small number of spots may reduce array effects, but can introduce considerable variation in the results. Centreing using the bulk of spots on the array is less variable. Similarly, normalisation methods based on regression methods can introduce variability. Except for normalisation methods based on spiking controls, all normalisation requires that most genes do not differentially express. Methods based on spatial location and/or intensity also require that the nondifferentially expressing genes are at random with respect to location and intensity. Spotting designs should be carefully done so that spot replicates are widely spaced on the array, and genes with similar expression patterns are not clustered. DISCUSSION: The tools for statistical design of experiments can be applied to microarray experiments to improve both efficiency and validity of the studies. Given the high cost of microarray experiments, the benefits of statistical input prior to running the experiment cannot be over-emphasised.     

An integrated bioinformatics approach to improve two-color microarray quality-control: impact on biological conclusions

Omics technology used for large-scale measurements of gene expression is rapidly evolving. This work pointed out the need of an extensive bioinformatics analyses for array quality assessment before and after gene expression clustering and pathway analysis. A study focused on the effect of red wine polyphenols on rat colon mucosa was used to test the impact of quality control and normalisation steps on the biological conclusions. The integration of data visualization, pathway analysis and clustering revealed an artifact problem that was solved with an adapted normalisation. We propose a possible point to point standard analysis procedure, based on a combination of clustering and data visualization for the analysis of microarray data.     

A comparison of parametric and nonparametric methods for normalising cDNA microarray data

Normalisation is an essential first step in the analysis of most cDNA microarray data, to correct for effects arising from imperfections in the technology. Loess smoothing is commonly used to correct for trends in log-ratio data. However, parametric models, such as the additive plus multiplicative variance model, have been preferred for scale normalisation, though the variance structure of microarray data may be of a more complex nature than can be accommodated by a parametric model. We propose a new nonparametric approach that incorporates location and scale normalisation simultaneously using a Generalised Additive Model for Location, Scale and Shape (GAMLSS, Rigby and Stasinopoulos, 2005, Applied Statistics, 54, 507-554). We compare its performance in inferring differential expression with Huber et al.'s (2002, Bioinformatics, 18, 96-104) arsinh variance stabilising transformation (AVST) using real and simulated data. We show GAMLSS to be as powerful as AVST when the parametric model is correct, and more powerful when the model is wrong.     

Background correction for cDNA microarray images using the TV+L1 model

MOTIVATION: Background correction is an important preprocess in cDNA microarray data analysis. A variety of methods have been used for this purpose. However, many kinds of backgrounds, especially inhomogeneous ones, cannot be estimated correctly using any of the existing methods. In this paper, we propose the use of the TV+L1 model, which minimizes the total variation (TV) of the image subject to an L1-fidelity term, to correct background bias. We demonstrate its advantages over the existing methods by both analytically discussing its properties and numerically comparing it with morphological opening. RESULTS: Experimental results on both synthetic data and real microarray images demonstrate that the TV+L1 model gives the restored intensity that is closer to the true data than morphological opening. As a result, this method can serve an important role in the preprocessing of cDNA microarray data.     

MicroPreP: a cDNA microarray data pre-processing framework

The user-friendly MicroPreP framework was developed to transform raw intensity data from cDNA microarrays into high-quality data. The main features of this software are: LOWESS normalisation; merging of DNA microarray data from changing slide versions; outlier detection; and slide quality assessment.     

Technical evaluation of cDNA microarrays

The aims were to evaluate the common reference design approach in microarray experiments and to evaluate the technical performance and the normalisation of cDNA microarrays with a limited number of spots. Total RNA from 3 normal and 3 tumour sample biopsies were used for synthesis of amino-allyl labelled cRNA. Equal amounts of cRNA from all samples were mixed as reference. After conjugation of cRNA with fluorophores (Cy3/Cy5), each individual tumour cRNA was hybridised to a cDNA microarray together with reference cRNA, scanned and analysed. We show that our procedures for producing cDNA microarrays are reproducible. The concordance between duplicated spots and replicate hybridisation was found to be high. We have demonstrated that our cDNA microarrays are of a high technical quality. The majority of the cDNA microarrays had low local spot background levels. Despite the high background levels for some local spots, variation could be minimized by locally weighted scatter plot smooth normalisation (LOWESS), which we showed was also suitable for normalisation of cDNA microarrays with a limited number of probes.     

Comparative evaluation of laser-based microarray scanners

Laboratories use different laser-based scanners to scan microarray images. To assess whether results from different scanners are comparable, and thus whether data from different laboratories can be compared, we scanned the same microarray slide with three commercial scanners that use different imaging techniques. After the acquisition of the microarray images produced by the three scanners, the images were quantified using a single imaging software package and protocol. The results were compared, and we found that the data obtained from the three scanners were comparable and that the variations caused by the use of different instruments were negligible, in spite of the fact that the scanners were based on different optical imaging techniques.     

Antibody microarray-based profiling of complex specimens: systematic evaluation of labeling strategies

Antibody microarrays have often had limited success in detection of low abundant proteins in complex specimens. Signal amplification systems improve this situation, but still are quite laborious and expensive. However, the issue of sensitivity is more likely a matter of kinetically appropriate microarray design as demonstrated previously. Hence, we re-examined in this study the suitability of simple and inexpensive detection approaches for highly sensitive antibody microarray analysis. N-hydroxysuccinimidyl ester (NHS)- and Universal Linkage System (ULS)-based fluorescein and biotin labels used as tags for subsequent detection with anti-fluorescein and extravidin, respectively, as well as fluorescent dyes were applied for analysis of blood plasma. Parameters modifying strongly the performance of microarray detection such as labeling conditions, incubation time, concentrations of anti-fluorescein and extravidin and extent of protein labeling were analyzed and optimized in this study. Indirect detection strategies whether based on NHS- or ULS-chemistries strongly outperformed direct fluorescent labeling and enabled detection of low abundant cytokines with many dozen-fold signal-to-noise ratios. Finally, particularly sensitive detection chemistry was applied to monitoring cytokine production of stimulated peripheral T cells. Microarray data were in accord with quantitative cytokine levels measured by ELISA and Luminex, demonstrating comparable reliability and femtomolar range sensitivity of the established microarray approach.