Table 2.

Predictive performances of RF models (quantified by the mean ± standard deviation values of AUC) trained and tested with different predictor sets in the ClinSeq®; and FHS cohort data

Predictors	Optimal # markers	Optimal AUC	p-value
CLIN	3 (ClinSeq®;), 3 (FHS)	0.69 ±0.02 (ClinSeq®;), 0.61 ±0.02 (FHS)	0.015 (ClinSeq®;), 0.080 (FHS)
SNP Set-2	21 (ClinSeq)®;, 21 (FHS)	0.99 ±0.01 (ClinSeq®;), 0.85 ±0.02 (FHS)	<0.001 (ClinSeq®;), <0.001 (FHS)
CLIN+SNP Set-2	21 (ClinSeq®;), 18 (FHS)	0.99 ±0.01 (ClinSeq®;), 0.83 ±0.01 (FHS)	<0.001 (ClinSeq®;), <0.001 (FHS)

“CLIN” corresponds to the nine clinical variables listed in Additional file 1: Table S1 (all variables except age and gender)