Quantifying similarity between motifs

Table 1 Mean ROC scores for various motif column comparison functions and score combination methods

Ranking method	ALLR	PCC	PCST	FIET	KLD	ED	SW
Sum	0.9823^a	0.9845^a	0.9786	0.9834	0.9793	0.9886	0.9809
AM	0.9595	0.9685	0.9619	0.9662	0.9736	0.9779	0.9735
GM	0.9643	0.9670	0.9630	0.9717	0.9724	0.9776	0.9720
p value	0.9786	0.9835	0.9797^a	0.9842^a	0.9864^a	0.9889^a	0.9861^a

The table reports the performance of the seven different column comparison functions using four different methods for combining scores: summing the raw scores, computing the arithmetic mean (AM), or computing the geometric mean (GM). Each entry is the mean receiver operating characteristic (ROC) score across all queries in the simulation. The table reports results for the S/8 sampling rate. ^aHighest ROC in the column. ALLR, average log-likelihood ratio; ED, euclidean distance; FIET, Fisher-Irwin exact test; KLD, Kullback-Leibler divergence; PCC, Pearson correlation coefficient; PCST, Pearson χ² test; SW, Sandelin-Wasserman function.

ISSN: 1474-760X