Identifying and mitigating batch effects in whole genome sequencing data

BMC Bioinformatics

Table 2 Retaining confirmed AMD associations in a candidate SNP analysis when batch is completely confounded with AMD status

CHR	Position^a	p-value	Percent missing	GQ20M05	GQ20M30	Diff GQ	LD corrected p-value
1	196,710,325	0.002199	4.928	NF	NF	NF	0.002122
3	64,719,689	NS	5.027	F	NF	NF	NS
3	99,762,695	NS	5.027	F	NF	NF	NS
6	43,858,890	NS	6.57	F	NF	F	NS
6	116,122,572	NS	6.471	F	NF	NF	NS
8	23,225,458	NS	6.72	F	NF	NF	NS
9	99,146,083	NS	5.475	F	NF	NF	NS
10	122,454,932	1.60E-05	0.6969	NF	NF	NF	1.79E-05
13	31,245,188	NS	5.625	F	NF	NF	NS
14	68,318,360	NS	5.226	F	NF	NF	NS
15	58,396,268	NS	5.625	F	NF	NF	NS
16	56,963,321	NS	3.833	NF	NF	F	NS
19	6,718,376	NS	3.534	NF	NF	NF	NS
19	44,919,689	2.30E-21^b	6.67	F	NF	NF	3.16E-21
22	32,663,679	NS	6.521	F	NF	NF	NS
22	38,080,269	NS	6.272	F	NF	NF	NS

NF is not filtered, F is filtered, GQ20M05 filter, filter sites with more than 5% missingness after setting genotypes with GQ < 20 to missing; GQ20M30 filter, filter sites with more than 30% missingness after setting genotypes with GQ < 20 to missing
Diff GQ, differential genotype quality filter, LD linkage disequilibrium, NS not significant in candidate SNP analysis at Bonferroni adjusted significance level: 0.05/16 = 0.00312
^aSites are reported in GRCh38 coordinates
^bWe detect APOE because our controls are enriched for Alzheimer’s cases

ISSN: 1471-2105