MiRNA array data analysis. Data preprocessing included normexp background correction (Ritchie et al., 2007) and print-tip loess normalization(Smyth and Speed, 2003). Statistical differences between groups was performed using the Bioconductor package limma (Smyth et al., 2005) using an empirical Bayes approach. Since each probe was spotted in duplicate on the arrays, within-array replication was assessed using the limmaduplicate Correlation function (Smyth et al., 2005). False discovery rate was assessed using the method of Benjamini and Hochberg (Benjamini and Hochberg, 1995). The raw data has been deposited in MIAME compliant Longhorn Array Database (LAD) (Killion et al., 2003) with the access number in progress.
