The objective of the proposed research is biomarker exploration for patients with Treatment Resistant Depression (TRD) undergoing subcallosal cingulate (SCC)-Deep Brain Stimulation (DBS). While SCC-DBS is a promising intervention for TRD, depression's heterogeneous nature, subjective clinical assessments, and slowly evolving symptom dynamics underscore the critical need for objective, brain-based biomarkers to reliably track and manage therapy. The first aim focuses on replicating a retrospective biomarker for tracking depression states in a new patient cohort using bilateral Local Field Potential (LFP) recordings from a next-generation DBS device, evaluating biomarker efficacy across recording conditions and hemispheric differences to distill information into specific frequency bands; suitable for newer DBS devices. The second aim develops a prospective biomarker for real-time depression state tracking using semi-supervised generative modeling to address distribution shifts across subjects and labeled data scarcity, enabling prospective tracking of slow depression dynamics while disentangling residual variations from true depression states. The third aim shifts focus to noninvasive electroencephalography (EEG)-captured evoked potentials during low frequency stimulation through DBS contacts, proposing latent modeling of short-timescale, stimulation-evoked EEG responses to classify treatment response status from single time-point measurements.