Computational study of associations between histone modification and protein-DNA binding in yeast genome by integrating diverse information

Table 1 Mean percentage of correct classifications and mean total number of correct classifications for 32 yeast TFs.

Rank	TF Name	Mean percentage of correct classifications in 10 test datasets	Mean total number of correct classifications in 10 test datasets
1	NRG1	90	29
2	FHL1	87	70
3	GAT3	83	9
4	RLM1	82	14
5	HIR1	80	9
6	GAL4	77	7
7	MSS11	75	7
8	SUM1	73	19
9	BAS1	72	13
10	GCN4	72	25
11	DAL81	71	12
12	NDD1	71	29
13	YAP1	71	15
14	MSN4	70	6
15	HAP4	69	17
16	HIR2	69	5
17	HSF1	69	16
18	SWI4	68	37
19	SWI5	68	27
20	YAP6	65	24
21	MBP1	64	31
22	MET31	64	9
23	ARG81	63	5
24	CAD1	63	11
25	ABF1	62	81
26	ACE2	60	17
27	FKH2	60	30
28	STE12	60	13
29	LEU3	56	5
30	STB1	56	6
31	MCM1	55	21
32	DIG1	51	7

The calculations are completed by applying Bayesian Neural Networks on 10 randomly selected test datasets (the half-available binding sites), which include diverse histone modifications (e.g. 3 histone acetylation, 5 histone methylation levels, nucleosome occupancy, and 2 histone acetyltransferase; Pokholok et al.); TF name with bold text represents yeast cell cycle related TFs [54], and TF rank order is based on mean percentage of correct classification in 10 test datasets.

ISSN: 1471-2164