Constructing gene expression-based diagnostic rules for understanding individualized etiology of heart failure

Abstract

Gene expression profiling has the potential to improve individualized etiology diagnosis. A statistical approach based on a multidimensional scaling (MDS) vector model is introduced to construct patient-specific rules for individualized diagnosis based on gene expression profiles. The method has a dual function of discovering new disease classes/subclasses as well as constructing patient-specific diagnostic rules without prior knowledge of class distinction. The diagnostic rule consists of two components: (1) diagnostic gene expression pattern that suggests a critical etiological condition associated with a disease category, and (2) patient-specific correlations to the diagnostic pattern. The method is applied to construct the diagnostic rule for heart failure by which the heart failure etiology has been successfully discerned with gene expression profiles. The diagnostic rule for two potential heart failure sub-classes has been constructed to further classify heart failure patients and exploit related molecular pathogenesis. Furthermore, the diagnostic gene expression patterns reveal molecular mechanisms relevant to heart failure, and facilitate biomarker identification. The method provides an approach to exploring feasibility of gene expression profiling in individualized etiology diagnosis for therapeutic decision-making.