Table 3. Multivariate analysis using ensemble tree models - Random forests approach results.

The random forest procedure was performed using 1000 trees on the same 49 variables obtained after univariate and multicollinearity analysis. The top important variables strongly associated with the outcome are listed based on positive variable importance (VIMP).

Variable	VIMP
Resting state right amygdala - right supplementary motor cortex	9.322006
rs3794834 biomarker	4.557822
Resting state right amygdala - left Heschl's gyrus	4.338526
NPY baseline assay	3.645515
rs1629089 biomarker	3.325563
rs8185002 biomarker	2.743513
Resting state right amygdala - left subcallosal cortex	2.585636
rs2008323 biomarker	2.421426
DTI right optic radiation white matter (FA)	2.110193
rs8099667 biomarker	1.925468
rs9903602 biomarker	1.740449
rs2282574 biomarker	1.692296
rs2839350 biomarker	1.580902
rs642506 biomarker	1.46731
rs40184 biomarker	0.983404
Resting mean SCL	0.753201
DTI right uncinate fasciculus white matter integrity (FA)	0.690579
rs12273539 biomarker	0.506212
DTI right superior longitudinal fasciculus white matter volume	0.071362