Table 1.

Mean cross-validated gene expression prediction performance of machine learning models in MESA populations

Population	EN	RF	SVR	KNN
AFA	0.0528	0.0041	0.0120	−0.0086
HIS	0.0479	0.0156	0.0289	0.0001
CAU	0.0596	0.0283	0.0437	0.0094
ALL	0.0733	0.0409	0.0476	0.0103

Elastic net (EN) had higher mean performance than each of the other machine learning models across the MESA populations (all paired t test p values < 4 × 10⁻¹⁹). AFA, MESA African American; CAU, MESA European American; HIS, MESA Hispanic American; ALL, all MESA; RF, random forest; SVR, support vector regression; KNN, K nearest neighbor.