Using OPLS-DA to find new hypotheses in vast amounts of gene expression data - studying the progression of cardiac hypertrophy in the heart of aorta ligated rat.

One of the great problems facing science today lies in data mining of the vast amount of data. In this study we explore a new way of using orthogonal partial least squares-discrimination analysis (OPLS-DA) to analyze multidimensional data. Myocardial tissues from aorta ligated and control rats (sacrificed at the acute, the adaptive and the stable phases of hypertrophy) were analyzed with whole genome microarray and OPLS-DA. Five functional gene transcript groups were found to show interesting clusters associated with the aorta ligated or the control animals. Clustering of "ECM and adhesion molecules" confirmed previous results found with traditional statistics. The clustering of "Fatty acid metabolism", "Glucose metabolism", "Mitochondria" and "Atherosclerosis" which are new results is hard to interpret, thereby being possible subject to new hypothesis formation. We propose that OPLS-DA is very useful in finding new results not found with traditional statistics, thereby presenting an easy way of creating new hypotheses.