The use of haplotype-specific transcripts improves sample annotation consistency

Biomarker Research

Table 3 Application of the score to public datasets

Individual	Sample description	CEL file ID	203290_at HLA-DQA1 *0401	213831_at HLA-DQA1 *0103	209728_at HLA-DRB4	HLA-Score	REDKX gender QC	Flag	Reason for flag	Decision
55	Time point 1	GSM155503.CEL	28	4182	48	010	F
	Time point 2	GSM155504.CEL	23	4606	72	010	F
	Time point 3	GSM155505.CEL	10	2412	24	010	F
	Time point 4	GSM155506.CEL	28	4765	21	010	F
54	Time point 1	GSM155499.CEL	33	516	985	011	F
	Time point 2	GSM155500.CEL	49	681	1245	011	F
	Time point 3	GSM155501.CEL	28	1073	3142	011	F
	Time point 4	GSM155502.CEL	26	914	2573	011	F
45	Time point 1	GSM155495.CEL	45	8041	187	011	F	HLA-score	Intensity of 209728_at HLA-DRB4 slightly above threshold	Not critical
	Time point 2	GSM155496.CEL	59	7619	123	010	F
	Time point 3	GSM155497.CEL	32	6385	105	010	F
	Time point 4	GSM155498.CEL	59	7062	40	010	F
35	Time point 1	GSM155475.CEL	1218	42	6444	101	F	HLA-score and REDKX QC-score	Gender different for patient samples	Possible sample mix-up; follow up
	Time point 2	GSM155476.CEL	436	1491	231	111	M
	Time point 3	GSM155477.CEL	508	1751	200	111	M
	Time point 4	GSM155478.CEL	126	113	5	000	M	HLA-score	Intensity of 209728_at HLA-DRB4 and 213831_at HLA-DQA1 10× to 46× smaller than those from other patient samples	Possible sample mix-up; follow up
32	Time point 1	GSM155471.CEL	1235	32	8231	101	F	HLA-score and REDKX QC-score
	Time point 2	GSM155472.CEL	420	1878	291	111	M		Gender different for patient samples	Possible sample mix-up; follow up
	Time point 3	GSM155473.CEL	1807	49	7332	101	F		Gender different for patient samples	Possible sample mix-up; follow up
	Time point 4	GSM155474.CEL	1128	42	7122	101	F

Each sample of the dataset is labelled with a three-digit score (one “1” or “0” flag per probe set). The presence of intra-score differences elicits further follow-up investigation as to the nature and source of the difference. Intensity thresholds mark an “on” or “off”-status of the transcript expression, flagged as “1” or “0”, respectively. The thresholds were empirically determined for each marker separately. In some instances, the score difference is due to mild threshold violation, in other instances it may be due to sample mix-ups. The score picks up those samples which are being flagged by the REDKX gender QC, and detects further samples with issues (labeled in bold). REDKX panel expression values are provided in Additional file 1: Table S1.

ISSN: 2050-7771