Independent component analysis, a new concept?
Signal Processing - Special issue on higher order statistics
Fast and robust fixed-point algorithms for independent component analysis
IEEE Transactions on Neural Networks
Comments on selected fundamental aspects of microarray analysis
Computational Biology and Chemistry
| Hi-index | 0.00 |
The number of statistical tools used to analyze transcriptome data is continuously increasing and no one, definitive method has so far emerged. There is a need for comparison and a number of different approaches has been taken to evaluate the effectiveness of the different statistical tools available for microarray analyses. In this paper, we describe a simple and efficient protocol to compare the reliability of different statistical tools available for microarray analyses. It exploits the fact that genes within an operon exhibit the same expression patterns. In order to compare the tools, the genes are ranked according to the most relevant criterion for each tool; for each tool we look at the number of different operons represented within the first twenty genes detected. We then look at the size of the interval within which we find the most significant genes belonging to each operon in question. This allows us to define and estimate the sensitivity and accuracy of each statistical tool. We have compared four statistical tools using Bacillus subtilis expression data: the analysis of variance (ANOVA), the principal component analysis (PCA), the independent component analysis (ICA) and the partial least square regression (PLS). Our results show ICA to be the most sensitive and accurate of the tools tested. In this article, we have used the protocol to compare statistical tools applied to the analysis of differential gene expression. However, it can also be applied without modification to compare the statistical tools developed for other types of transcriptome analyses, like the study of gene co-expression.