Table 1 Performance measures from the PRS pipeline validation study at the clinical laboratory

		Asthma	Atrial fibrillation	Breast cancer	Chronic kidney disease	Coronoary heart disease	Hypercholesterolemia	Obesity/BMI	Prostate cancer	Type 1 diabetes	Type 2 diabetes
PRS accuracy: Pearson correlation between PRS from array and WGS (%)		99.3	98.6	93.0	98.3	98.2	95.9	99.5	96.4	99.5	98.8
PRS precision: PRS pipeline repeatability (%)		100	100	100	100	100	100	100	100	100	100
PRS precision: PRS pipeline reproducibility (z-score standard deviation)		0.0020	0.0010	0.0040	0.0001	0.0010	0.0050	0.0020	0.0006	0.0001	0.0010
PRS site missingness (%)		0.69	1.20	0.32	0.69	0.46	1.20	0.70	2.97	2.97	0.70
Odds ratio (95% CI)	European	1.95 (1.43–2.65)	2.32 (2.07–2.61)	2.47 (2.20–2.77)	3.6 (3.11–4.17)	2.3 (2.07–2.56)	4.16 (2.59–6.44)		3.67 (3.57–3.76)	12.97 (7.29–20.40)	4.21 (3.66–4.84)
	African American	1.83 (1.24–2.70)	2.19 (1.38–3.38)	1.61 (1.38–1.87)	2.66 (2.01–3.51)	1.68 (1.39–2.03)	3.16 (1.92–5.01)		2.95 (2.60–3.30)	20.45 (10.77–38.83)	2.55 (2.09–3.11)
	Hispanic	3.12 (1.32–7.44)	2.27 (1.09–4.50)	2.05 (1.10–3.83)	4.93 (2.46–9.89)	2.16 (1.47–3.19)	4.02 (2.72–5.83)		n.d.	n.d.	6.87 (3.11–15.15)
	Asian	n.d.	n.d.	2.22 (1.99–2.47)	3.81 (1.91–7.59)	n.d.	3.75 (3.15–4.42)		n.d.	n.d.	4.58 (4.00–5.23)

PRS pipeline accuracy is assessed as the Pearson correlation between scores derived from polymerase chain reaction (PCR)-free 30X whole genome sequencing (WGS) and those derived from imputed genotyping data (GDA) in the same 70 specimens. Pearson correlation is shown in the mean correlation across all ancestry groups tested. PRS pipeline precision (repeatability) is the measure of concordance in PRSs calculated from the same 70 specimens, run through the pipeline ten times over the course of two weeks. PRS pipeline precision (reproducibility) is assessed using three samples, each run six times end-to-end and then compared in a pairwise manner. The z-score standard deviation is used as a measure of variability. PRS site missingness is the percentage of genomics sites in the original score that are missing from the final imputed dataset. Odds ratios for high PRS versus not high PRS are derived from the condition-specific cohorts and calculated by each condition group lead across the ancestries available. Odds ratio information for obesity/BMI is in preparation for publication by the Genetic Investigation of ANthropometric Traits consortium. BMI, body mass index; CI, confidence interval; n.d., no data.

Search