A unique pair of microarray datasets for microRNA profiling: data without careful design (Dataset B)

We set out to demonstrate the logistic feasibility of careful study design in microarray studies and the level of scientific benefits it can provide, in comparison with post-hoc data adjustment such as normalization, for the purpose of reducing the impact of handling effects on the accuracy and reproducibility of microarray data analysis. Towards this end, we conducted a study of microRNA expression using endometroid endometrial cancer tumors (n=96) and serous ovarian cancer tumors (n=96) that were all primary, untreated, and collected in 2000-2012 at Memorial Sloan Kettering Cancer Center. The same set of tumor tissue samples were profiled twice using the Agilent microRNA microarrays (Human miRNA V16.0), once with uniform handling and careful design for the allocation of arrays to samples and the second study without. In the first study, arrays were assigned to tumor samples using blocked randomization and they were processed by one experienced technician in one single run. In the second study, arrays were assigned to tumor samples in the order of sample collection and they were handled by two technicians (the first of whom handled the first dataset) in multiple batches (with each batch on a separate date), mimicking typical practice. Dataset B is from the second study. Tumor type: advanced serous ovarian cancer versus endometrioid endometrial cancer; batch: processing batches, Batch 1 to Batch 5.