RESUMO
BACKGROUND: Analysis of gene expression in peripheral blood samples is increasingly being applied in biomarker studies of disease diagnosis and prognosis. Although knowledge of interindividual and interethnic variation in gene expression is required to set ethnicity-specific reference intervals and to select reference genes and preferred markers from a list of candidate genes, few studies have attempted to characterize such biological variation on a genomewide scale. METHODS: The genomewide expression profiles of 11 355 transcripts expressed among 210 multiethnic individuals of the HapMap project were obtained and analyzed; 4 replicates were included for each sample. The total biological CV in gene expression (CV(b)) was partitioned into interindividual (CV(g)), inter-ethnic group (CV(e)), and residual components by random-effects mixed models. RESULTS: CV(g) was the major component of CV(b), and the differences among transcripts were large (up to 38%). Distinct groups of genes were characterized by CV values and expression levels. Of the genes with lowest biological variation (CV(b) < 1.5%), 35 genes were highly expressed, whereas 32 had intermediate or low expression. Although CV(g) was almost always greater than CV(e), we identified 10 genes in which ethnic variation predominated (range, 8%-18%). On the other hand, 17 annotated genes were highly variable with CV(g) values ranging between 15% and 38%. CONCLUSIONS: Genomewide analysis of gene expression variation demonstrated biological differences among transcripts. Transcripts with the least biological variation are better candidates for reference genes, whereas those with low interindividual variation may be good disease markers. The presence of interethnic variation suggests that ethnicity-specific reference intervals may be necessary.