The divergence test statistic by sample type. These box plots show the distributions of the divergence test statistic for IP-Input and Input-Input comparisons for the ENCODE training data grouped into sample types: (a) histone mark in cancer cells; (b) histone mark in normal cells; (c) TF in cancer cells; and (d) TF in normal cells. The distributions of the divergence test statistic have slightly stronger mixing for histone data (a,b) compared to TF data (c,d), as evidenced by the higher whiskers in the TF IP-Input box plots. Moreover, the Input-Input comparisons for cancer samples (a,c) show a greater number of outliers denoted by red plus signs with large divergence test values compared to normal samples (b,d). To account for these differences, CHANCE estimates false discovery rates separately for each sample type. TFBS, transcription factor binding site.
Diaz et al. Genome Biology 2012 13:R98 doi:10.1186/gb-2012-13-10-r98