<p>The amygdala and its cortical connections play an important role in antidepressant treatment of major depressive disorder (MDD). While it can be subdivided according to its cortical connections, the anatomical and molecular heterogeneity across these subdivisions in MDD patients responding to different antidepressants remain unclear. Therefore, we aimed to investigate whether nuclei-level cortico-amygdala white matter (WM) connectivity differentiates responses to two first-line treatments - selective serotonin reuptake inhibitor (SSRI) and serotonin-norepinephrine reuptake inhibitors (SNRI); and to clarify the molecular underpinnings of such difference. From diffusion tensor imaging data in a discovery cohort of 106 MDD patients who response to SSRI (SSRI-Improvers) or SNRI (SNRI-Improvers), cortico-amygdala WM connectivity was estimated via probabilistic tractography. K-modes clustering was applied to cluster the amygdala into subdivisions by cortical projections. Fractional anisotropy (FA) values of each cortico-subdivision WM connection were then compared between SSRI-Improvers and SNRI-Improvers. Finally, the discriminative WM connection was decoded using microscale transcriptomic and chemo-architecture analysis, and its relevance was validated in an independent replication cohort of 66 MDD patients through targeted single-nucleotide polymorphisms (SNP) genotyping. A key difference in WM connection between frontal lobe and medial amygdala nucleus (medialAMY) was identified, which refers to a pharmacological propensity pattern. Such pattern was aligned with the serotonin and noradrenaline neurotransmitter system in a group-specific manner, and mediated the effects of genetic risks of <i>SLC6A4</i> on treatment inclination. The convergence of macroscale pharmacological propensity connectome, microscale molecular architecture, and genetic variation provides new testable evidence to inform decision-making in antidepressant treatment of MDD.</p>