Total raw reads were filtered by sequencing quality. For every read, the number of quality values that were below 8 was counted. The reads were removed if the numbers were over 9 or 14 for 35mer or 50mer single end reads, respectively. Noncoding RNAs (mostly rRNAs and transfer RNAs) and adaptors were filtered out using Mapreads mapper (ABI, Carlsbad, CA). Two mismatches out of 20mer reads were allowed for the mapper. The filtered sequencing data were modified to .fq files using fq_all2std.pl program (http://maq.sourceforge.net/fq_all2std.pl), removing reads that contain quality values of -1. For each sample, TopHat mapping was performed against the human reference genome (hg18) using default options (20). To visualize the mapping data, .sam files were modified using igvtools (http://www.broadinstitute.org/software/igv/download) and visualized using the Integrative Genomics Viewer (IGV) (http://www.broadinstitute.org/software/igv/) including another TopHat output (.wig file).
