Circulating miRNA microarray data from breast cancer patients

Independent studies have reported that circulating miRNAs have the potential as biomarkers; however, no consolidated guidelines for the discovery process are available. We developed a pipeline using innovative applications of existing bioinformatics methods to (1) face the inapplicability of the classical normalization methods, (2) detect global differences of miRNA distributions between the comparison classes and (3) develop a 'robust’ classifier. 78 samples (26 hemolyzed and 52 non hemolyzed) were 1:2 paired with caliper matching according to disease status, age at drawing and drawing year, starting from a total of 208 hybridized samples collected between 1987 and 2004 (including 94 paired samples according to tumor ER status, age at drawing, drawing year, time from surgery, 7 additional samples from 'no evidence of disease' (NED) patients, 10 references, 1 healthy donor and 2 replicated samples)