Boosting tissue-specific prediction of active cis-regulatory regions through deep learning and Bayesian optimization techniques

Table 1 For each dataset (on columns) we report the number of samples per class, as well as the cardinality of the dataset composed solely by enhancers and promoters (rows “Total E+P”) for genome version hg19 (Top table) and genome version hg38 (Bottom table)

Genome version	Labels	HepG2	K562	GM12878	Total	HelaS3
hg19	Active enhancer (AE)	1465	894	2878	5237	1847
	Inactive enhancer (IE)	34,556	34,392	28,156	97,104	32,179
	Active promoter (AP)	11,467	10,076	10,816	32,359	10,759
	Inactive promoter (IP)	96,184	82,829	73,891	252,904	79,004
	Total E + P	143,672	128,191	115,741	387,604	123,789
	Active exon (AX)	9931	9033	8226	9123
	Inactive exon (IX)	19,071	20,261	19,078	22,071
	Unknown (UK)	79,417	78,081	80,004	81,502
	Total	25,209	235,566	223,049	236,485
hg38	Active enhancer (AE)	7177	5524	11,589	24,290
	Inactive enhancer (IE)	56,108	57761	51,696	165,565
	Active promoter (AP)	14,092	12,524	14,036	40,652
	Inactive promoter (IP)	85,789	87,357	85,845	258,991
	Total E + P	163,166	163,166	163,166	489,498

Column “Total” allows comparing the total cardinality of CRRs across the hg19 and the hg38-datasets. Since we also have non-CRRs regions for genome version hg19, row “Total” in the top table reports the total number of samples per cell line in the hg19 dataset

ISSN: 1471-2105