Biostatistical considerations of the use of genomic DNA reference in microarrays

Using genomic DNA as common reference in microarray experiments has recently been tested by different laboratories (2, 3, 5, 7, 9, 20, 24-26). While some reported that experimental results of microarrays using genomic DNA reference conformed nicely to those obtained by cDNA: cDNA co-hybridization method, others acquired poor results. We hypothesized that these conflicting reports could be resolved by biostatistical analyses. To test it, microarray experiments were performed in a γ-proteobacterium Shewanella oneidensis. Pair-wise comparison of three experimental conditions was obtained either by direct cDNA: cDNA co-hybridization, or by indirect calculation through a Shewanella genomic DNA reference. Several major biostatistical techniques were exploited to reduce the amount of inconsistency between both methods and the results were assessed. We discovered that imposing the constraint of minimal number of replicates, logarithmic transformation and random error analyses significantly improved the data quality. These findings could potentially serve as guidelines for microarray data analysis using genomic DNA as reference.