A computational approach to microarray universal reference sample

The choice of a reference material in two-colour microarray experiments is an important issue of the experimental design. We consider a number of cell lines derived from a variety of primary tissues in order to construct a reference material. The aim of the present study is to understand the correlation structure of the data and develop a common reference that would enable comparison of expression levels in multiple microarray hybridizations with high coverage. We analyze 22 cultured cell line samples using a common reference pool. After estimating the coverage for each cell line or combinations of cell lines based on methods suggested by the literature, we employ stochastic optimization techniques to estimate the optimal set of cell lines in the reference sample. We found that only a subset of cell lines is necessary to achieve coverage as high or higher than that of the original reference sample used. We tested experimentally the performance of the new reference sample suggested and found that its coverage outperforms the coverage achieved by Stratagenepsilas human reference sample for the particular platform.