To obtain insights into the molecular pathogenesis of heart failure in humans, we have analyzed the expression profiles of >12,000 genes in a total of 17 human specimens of right atrial myocytes. From this large data set, we here tried to identify gene clusters, expression level of which is correlated precisely with clinical parameter values of cardiac function. We could reveal that cardiac myocytes with normal sinus rhythm were clearly differentiated, in the point of view of gene expression, from those with atrial fibrillation. Further, an expression profile-based prediction of arrhythmia by a newly developed "weighted-distance method" could efficiently diagnose our samples. We could even construct calculation formulae for the values of left ventricular ejection fraction based on the expression level of selected genes. To our best knowledge, this is the first report to indicate that pumping ability of heart can be predicted by any measures of atrium. (C) 2004 Elsevier Inc. All rights reserved.