Extended Data Fig. 1: Overview of methods: data and cluster-based analyses.

Data, Patients’ raw CES-D scores were split into three categories: an absence of (≤6; ADS), intermediate (7-22) and severe (≥23; SDS) depressive symptoms. We tested which patient and tumor characteristics were significantly associated with the CES-D category as core model using a Bayesian categorical multiple regression. Structural MRI scans were utilized to create semi-automated segmentations followed by a non-linear registration to MNI152 standard space. With the segmentations we constructed a) tumor distribution maps of all patients and b) patient-specific disconnectomes and disconnectome distribution maps of all patients. Cluster-based analyses. Patients’ tumor segmentations (T) and continuous CES-D scores were used as input for lesion to symptom mapping with sparse canonical correlation analysis. This resulted in a tumor symptom map with clusters of voxels corresponding to either severe or an absence of depressive symptoms. Accordingly, disconnectomes and continuous CES-D scores were used as input for a disconnectome symptom map.