Accounting for overlapping annotations in genomic prediction models of complex traits

BMC Bioinformatics

Table 2 Average rankings of large QTLs by estimated posterior variance

Annotation scenario		None	A (2 annot.)			B (4 annot.)			C (4 annot.)			D (9 annot.)
h²	k_large (%)	R	RC	RC\(\pi\)	RC+	RC	RC\(\pi\)	RC+	RC	RC\(\pi\)	RC+	RC	RC\(\pi\)	RC+
0.2	1	10286	501	479	342	652	617	433	4930	4446	4212	1268	1112	322
	2.5	2991	140	120	91	188	167	110	1577	1933	1755	340	303	93
	5	597	21	19	14	29	25	18	392	361	425	44	41	16
0.5	1	2711	162	140	88	194	152	102	1482	1303	1436	365	261	102
	2.5	127	12	11	8	21	12	10	73	74	104	23	20	10
	5	4	3	3	3	3	3	3	4	26	55	3	3	3

Mean rank (by decreasing estimated posterior variance) of large QTLs, averaged across 50 independent datasets for each setting (heritability, \(k_\text {large}\)) and each method (R = BayesR, RC = BayesRC, RC\(\pi\) = BayesRC\(\pi\) and RC+ = BayesRC+). With the exception of BayesR, which does not use annotations, results are presented by annotation scenario (A, B, C and D). Boldface is used to indicate the best ranking obtained for each setting. As each dataset contains 5 large QTLs, the highest average ranking that can be obtained is equal to 3 (average of 1 to 5)

ISSN: 1471-2105