E-learning in analysis of genomic and proteomic data 2. Data analysis 2.2. Analysis of high-density genomic data 2.2.1. DNA microarrays 2.2.1.10. Meta-analysis of microarray experiments

authors: Ivana Ihnatová, Eva Budinská

Basically, the analysis of microarrays can be divided in four levels. First level is represented by analysis of image and obtaining numerical values of intensities for each spot and channel. Then, the process of searching for differentially expressed genes follows. These genes are then used in the third level of analysis for class comparison, prediction or discovery. Analysis at fourth level is called meta-analysis, and here the integration of results of individual experiments is performed.

In praxis we face the problem of different results of two microarray studies with the same aim performed under the similar experimental conditions. Usually, the lists of selected significant genes have different length, contain different genes or the same genes occur at different position.

There are several causes of these differences occurring during the analysis of microarrays. First and very important reason is the presence of noise in microarray data. Actually each part of microarray experiment is a source of noise. During the analysis of image especially during quality control suspicious spots are excluded from further analysis. It brings missing values into data. Additionally, the biggest advantage of microarray – the ability to measure thousands of parameters (genes) at once is with small number of slides makes statistical analysis of such a data difficult. Other differences are caused by different experimental conditions, including different microarray platforms, different laboratory protocols or different samples.

A solution to these differences between individual microarray studies seems to be a meta-analysis of microarrays. By meta-analysis we mean analysis of published results. Its main advantage is the enhancement of statistical power for discovery of differentially expressed genes. However, meta-analysis itself has to deal with problems arising from combining of input datasets. This leads to a question whether it is possible to compare microarray studies or not. This question has not been definitely solved yet. There are authors (Choi et al.,2004) that advocate combining datasets from different platforms, others (Kuo et al.,2002, Jarvinen et al., 2004)] do not recommend it.