Deep brain stimulation (DBS) can be an effective therapy in patients with severe, treatment-refractory Tourette syndrome (TS) for improving tics, comorbid obsessive-compulsive behavior (OCB), and quality of life. However, clinical outcomes are variable; some patients experience life-changing improvements and others experience no effects. One contributor to the variability in clinical outcomes is that the applied stimulation varies across patients, which means that different neuroanatomical structures or networks may be modulated in different patients. Despite this variability, the brain regions, fiber pathways, or distributed networks associated with improvement in tics or comorbid OCB are unknown. As a result, no current methods exist to predict symptom improvement based on the applied stimulation, and it is unclear how to target stimulation most effectively. Progress toward identifying predictors of the therapeutic response has been hindered by small cohort sizes in research studies, since relatively few patients undergo DBS for TS at individual clinical centers. In this dissertation, we curated a multicenter, retrospective dataset of a large cohort of patients who have undergone DBS for TS. Image-based computational models were constructed for each patient to estimate the effects of DBS on local neuroanatomical structures and distributed networks. Our objectives were to identify the local brain regions, fiber pathways, and distributed networks modulated by DBS that were associated with symptom improvement and evaluate whether they could be used to predict clinical outcomes. The results showed that the applied stimulation was variable across patients, but there was no clear relationship between the local brain regions modulated by DBS and symptom improvement. In contrast, stimulation-dependent structural connectivity to distributed cortico-basal ganglia-thalamo-cortical networks and activation of specific local basal ganglia fiber pathways were associated with improvement in tics and OCB. Modulation of the identified networks and fiber pathways was predictive of clinical outcomes across patients. Our results could guide data-driven approaches to targeting DBS and other neuromodulation therapies in individual patients with TS and generate initial hypotheses about the underlying therapeutic mechanisms. Collectively, this dissertation is an essential step toward improving DBS therapy to reliably alleviate symptoms and improve the quality of life for patients with severe, treatment-refractory TS.