Table 2.

Model quality measures (sensitivity, specificity, Cohen’s κ, accuracy, AUC curve and Brier score) for the elastic net machine learning model, Reese et al. cord blood, Richmond et al. 568 CpG, Richmond et al. 19 CpG score the gradient boosting machine, random forest and support vector machine models that were among the four best performing models in our analysis. Results provided in this table are based on the Raine Study test data ($n=198$), NFBC1986 ($n=478$), and NFBC1966 ($n=602$).

	Sensitivity	Specificity	Cohen’s κ	Accuracy	AUC	Brier score	# CpGs required
Raine Study test data set
Elastic net score	0.91	0.76	0.68	0.83	0.87	0.13	204
Gradient boosting machine	0.91	0.82	0.72	0.88	0.88	0.1	1,511
Random forest	0.87	0.73	0.58	0.83	0.83	0.17	1,511
Support vector machine	0.87	0.73	0.6	0.83	0.85	0.13	1,511
Reese score	0.88	0.72	0.6	0.83	0.85	0.21	28
Richmond score 568 CpGs	0.7	0.68	0.34	0.69	0.72	0.22	568
Richmond score 19 CpGs	0.79	0.58	0.37	0.72	0.73	0.22	19
NFBC1986
Elastic net score	0.87	0.75	0.56	0.84	0.85	0.13	204
Gradient boosting machine	0.95	0.29	0.19	0.54	0.74	0.39	1,511
Random forest	0.79	0.16	0.06	0.64	0.54	0.24	1,511
Support vector machine	0.87	0.44	0.33	0.77	0.79	0.16	1,511
Reese score	0.87	0.61	0.46	0.82	0.8	0.18	28
Richmond score 568 CpGs	0.65	0.76	0.34	0.74	0.71	0.22	568
Richmond score 19 CpGs	0.65	0.77	0.31	0.68	0.73	0.22	19
NFBC1966
Elastic net score	0.72	0.78	0.39	0.73	0.8	0.19	204
Gradient boosting machine	0.88	0.26	0.1	0.45	0.68	0.48	1,511
Random forest	0.77	0.18	0.05	0.64	0.48	0.24	1,511
Support vector machine	0.88	0.45	0.33	0.76	0.75	0.2	1,511
Reese score	0.72	0.7	0.32	0.71	0.73	0.18	28
Richmond score 568 CpGs	0.66	0.63	0.22	0.69	0.72	0.22	568
Richmond score 19 CpGs	0.61	0.72	0.23	0.63	0.73	0.22	19

Note: AUC, area under the receiver operator curve.