. 2020 Oct 29;143(12):3763–3775. doi: 10.1093/brain/awaa334

Table 2.

Our novel cortex clock outperforms existing DNAm age algorithms in human cortex samples

	Testing dataset (n = 350)				Independent test dataset (n = 1221)
	DNAmClock_Cortical	DNAmClock_Multi	DNAmClock_Blood	DNAmClock_Pheno	DNAmClock_Cortical	DNAmClock_Multi	DNAmClock_Blood	DNAmClock_Pheno
Correlation (r)	0.99	0.96	0.95	0.8	0.83	0.65	0.52	0.32
RMSE, years	3.58	9.52	18.86	60.16	5.12	20.12	33.46	82.28

Accuracy statistics between DNAm age estimates and chronological age using our DNAmClockCortical, Horvath’s multi-tissue clock (DNAmClock_Multi) (Horvath, 2013), Zhang’s elastic net blood clock (DNAmClock_Blood) (Zhang et al., 2019) and Levine’s Pheno Age clock (DNAmClock_Pheno) (Levine et al., 2018) in both the testing (n = 350 cortical samples) and the independent test (n = 1221 cortical samples) datasets. RMSE = root mean squared error.