To investigate the correlation between differentially expressed probes and cellular composition, we performed hierarchical clustering on the 374 differentially expressed probes. For each cluster of probes, we calculated spearman correlations between each probe by averaging the expression measurements of the two visits of the PAX samples and cell count parameters (neutrophils, lymphocytes and monocytes, mean cell volume and hemoglobin concentration). Subsequently, we compared the mean spearman correlations of the probes in each cluster with mean spearman correlation of all detected probes excluding the differentially expressed probes using a Wilcoxon rank-sum test. Multivariate analyses were performed using Evince (UmBio). All other analyses were conducted within the statistical scripting environment, R v2.4.1 [31].
