“…The majority of ontological analysis methods rely on threshold-dependent, discrete statistical procedures, such as Pearson's chi-square or Fisher's exact tests, to test for the relative enrichment of gene categories with in lists of significantly differentially expressed genes (reviewed in Khatri and Draghici, 2005). Recently, however, some authors have proposed threshold-free methods based on the Kolmogorov-Smirnov (KS) as an attractive alternative to the discrete approaches (Ben-Shaul et al, 2005;Barry et al, 2005;Lamb et al, 2003;Mootha et al, 2003;Nilsson et al, 2007). In essence, KS-based analysis amounts to computing the empirical distributions of gene-specific differential expression scores (usually t-statistics or varianceregularized variants thereof (Cui et al, 2005;Smyth, 2004;Storey and Tibshirani, 2003) for the gene category being assessed and for a reference gene population (usually all genes on the array or the category complement).…”