Table of Contents Author Guidelines Submit a Manuscript
Comparative and Functional Genomics
Volume 4 (2003), Issue 3, Pages 300-317
http://dx.doi.org/10.1002/cfg.298
Primary research paper

Reproducibility Assessment of Independent Component Analysis of Expression Ratios From DNA Microarrays

1Department of Genetics, University of Cambridge, Downing Street, Cambridge CB2 3EH, UK
2Inference Group, Cavendish Laboratory, Madingley Road, Cambridge CB3 0HE, UK

Received 6 November 2002; Revised 30 March 2003; Accepted 4 April 2003

Copyright © 2003 Hindawi Publishing Corporation. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

DNA microarrays allow the measurement of transcript abundances for thousands of genes in parallel. Most commonly, a particular sample of interest is studied next to a neutral control, examining relative changes (ratios). Independent component analysis (ICA) is a promising modern method for the analysis of such experiments. The condition of ICA algorithms can, however, depend on the characteristics of the data examined, making algorithm properties such as robustness specific to the given application domain. To address the lack of studies examining the robustness of ICA applied to microarray measurements, we report on the stability of variational Bayesian ICA in this domain. Microarray data are usually preprocessed and transformed. Hence we first examined alternative transforms and data selections for the smallest modelling reconstruction errors. Log-ratio data are reconstructed better than non-transformed ratio data by our linear model with a Gaussian error term. To compare ICA results we must allow for ICA invariance under rescaling and permutation of the extracted signatures, which hold the loadings of the original variables (gene transcript ratios) on particular latent variables. We introduced a method to optimally match corresponding signatures between sets of results. The stability of signatures was then examined after (1) repetition of the same analysis run with different random number generator seeds, and (2) repetition of the analysis with partial data sets. The effects of both dropping a proportion of the gene transcript ratios and dropping measurements for several samples have been studied. In summary, signatures with a high relative data power were very likely to be retained, resulting in an overall stability of the analyses. Our analysis of 63 yeast wildtype vs. wild-type experiments, moreover, yielded 10 reliably identified signatures, demonstrating that the variance observed is not just noise.