Table S5.

Exported feature set performance (similar platform).

Feature set similar platforms	RF			R-SVM			S-SVM			Across
Method	Sen	Spe	bAcc	Sen	Spe	bAcc	Sen	Spe	bAcc	Sen	Spe	bAcc
MG	71	46	59	68	69	69	73	72	73	71	62	67
SG	76	39	58	67	71	69	71	71	71	71	60	66

Notes: The table shows the mean sensitivity, specificity and balanced accuracies for external validation of feature sets covering the following validation: Feature sets defined by AF1 and validated in AF2 and AF3. Feature sets defined by AF2 and validated in AF3 and vice versa, using either metagenes (MG) or single genes (SG) as features and using random forest (RF), support vector machine with a radial-based kernel (R-SVM) or a sigmoid kernel (S-SVM). Across shows the mean of the results across the three classification methods.