Promoter features related to tissue specificity as measured by Shannon entropy

Table 5 The most significant indicators of the degree of tissue specificity: start CpG island, TATA box, and YY1 site

Features			Total fraction	H 0-3	H 3-4	H 4-5
CGI	TATA	YY1		Most specific	Semi-specific	Least specific
			3,552	271	602	2679
			1.00	0.08	0.17	0.75
CGI+			2,434	56	306	2072
			0.69	0.02	0.13	0.85
				0.30	0.74	1.13
CGI-			1,118	215	296	607
			0.31	0.19	0.26	0.54
				2.52	1.56	0.72
	TATA+		604	136	175	293
			0.17	0.23	0.29	0.49
				2.95	1.71	0.64
	TATA-		2,949	135	427	2,387
			0.83	0.05	0.14	0.81
				0.60	0.85	1.07
CGI+	TATA+		284	19	82	183
			0.08	0.07	0.29	0.64
				0.88	1.70	0.85
CGI-	TATA+		320	117	93	110
			0.09	0.37	0.29	0.34
				4.79	1.71	0.46
CGI+	TATA-		2,150	37	224	1,889
			0.61	0.02	0.10	0.88
				0.23	0.61	1.16
CGI-	TATA-		798	98	203	497
			0.22	0.12	0.25	0.62
				1.61	1.50	0.83
		YY1+	293	1	16	276
			0.08	0.00	0.05	0.94
				0.04	0.32	1.25
CGI+		YY1+	261	1	10	250
			0.07	0.00	0.04	0.96
				0.05	0.23	1.27
CGI+		YY1-	2,173	55	296	1,822
			0.61	0.03	0.14	0.84
				0.33	0.80	1.11
CGI-		YY1-	1,086	215	290	581
			0.31	0.20	0.27	0.53
				2.59	1.58	0.71
CGI-		YY1+	32	0	6	26
			0.01	0.00	0.19	0.81
				0.00	1.11	1.08

The three columns on the left indicate the combination of features considered; empty cells indicate that the feature is not considered. The 'Total fraction' column indicates the number of promoters with each feature combination (in bold) and the corresponding fraction of all genes considered. The three columns on the right give statistics for matching genes in three bands of tissue specificity. The top two lines give the number and corresponding fraction of all genes considered for each band. For each feature combination, the numbers indicate the number (top, bold), fraction (middle), and enrichment ratio (bottom) of matching genes. The enrichment ratio is the fraction of promoters of genes in the entropy band that contain a feature divided by the band's fraction among all genes considered. For example, specific genes are best recognized by a combination of TATA box (TATA+) and lack of a CpG island (CGI-), which enriches the fraction of such genes from 8% to 37% - a factor of 4.79. Nonspecific genes are most specifically recognized by CpG islands and YY1 sites, which returns a set that is 96% nonspecific genes, but only matches 7%/75% = 10% of the nonspecific genes.

ISSN: 1474-760X