Independent component analysis of Alzheimer's DNA microarray gene expression data

Abstract

ABSTRACT: BACKGROUND: Gene microarray technology is an effective toolto investigate the simultaneous activity of multiple cellular pathwaysfrom hundreds to thousands of genes. However, because data in thecolossal amounts generated by DNA microarray technology are usuallycomplex, noisy, high-dimensional, and often hindered by low statisticalpower, their exploitation is difficult. To overcome these problems,two kinds of unsupervised analysis methods for microarray data: principalcomponent analysis (PCA) and independent component analysis (ICA)have been developed to accomplish the task. PCA projects the datainto a new space spanned by the principal components that are mutuallyorthonormal to each other. The constraint of mutual orthogonalityand second-order statistics technique within PCA algorithms, however,may not be applied to the biological systems studied. Extractingand characterizing the most informative features of the biologicalsignals, however, require higher-order statistics. RESULTS: ICA isone of the unsupervised algorithms that can extract higher-orderstatistical structures from data and has been applied to DNA microarraygene expression data analysis. We performed FastICA method on DNAmicroarray gene expression data from Alzheimer's disease (AD) hippocampaltissue samples and consequential gene clustering. Experimental resultsshowed that the ICA method can improve the clustering results ofAD samples and identify significant genes. More than 50 significantgenes with high expression levels in severe AD were extracted, representingimmunity-related protein, metal-related protein, membrane protein,lipoprotein, neuropeptide, cytoskeleton protein, cellular bindingprotein, and ribosomal protein. Within the aforementioned categories,our method also found 37 significant genes with low expression levels.Moreover, it is worth noting that some oncogenes and phosphorylation-relatedproteins are expressed in low levels. In comparison to the PCA andsupport vector machine recursive feature elimination (SVM-RFE) methods,which are widely used in microarray data analysis, ICA can identifymore AD-related genes. Furthermore, we have validated and identifiedmany genes that are associated with AD pathogenesis. CONCLUSION:We demonstrated that ICA exploits higher-order statistics to identifygene expression profiles as linear combinations of elementary expressionpatterns that lead to the construction of potential AD-related pathogenicpathways. Our computing results also validated that the ICA modeloutperformed PCA and the SVM-RFE method. This report shows that ICAas a microarray data analysis tool can help us to elucidate the moleculartaxonomy of AD and other multifactorial and polygenic complex diseases.