Gradient Biotech

Cohort comparison

Compare brain connectivity across participant groups using batch graph metrics, Network-Based Statistics (NBS), and group-level metric comparison.

Research question

Which connectivity edges or graph-theoretic properties differ significantly between patient and control groups, or between treatment conditions?

Who this is for

  • Psychiatry and neurology research groups studying disorder-related dysconnectivity
  • Clinical trial teams measuring treatment effects on brain network endpoints
  • Cognitive neuroscience labs comparing task or resting-state connectivity across conditions

Data requirements

DataRequiredPurpose
Connectivity matrix per subjectYesNBS and batch graph metrics input
Group labels on datasetsYesDefine comparison groups
Subject metadata CSVNoDiagnosis, cognitive scores, medications for report context

Minimum cohort size: at least two groups with multiple datasets each for meaningful NBS permutation testing.

Workflow

Upload matrices for all subjects → Assign group labels → Batch graph metrics → NBS → Compare metrics → Interpret

Step 1 — Upload cohort datasets

Create a study and upload connectivity matrices for all participants. Matrices can come from precomputed connectomes, fMRI functional connectivity pipelines, or EEG-derived PLV matrices.

Step 2 — Import subjects (optional)

Import a subjects CSV for participant metadata:

external_id,age,sex,group_label,diagnosis
S001,34,F,control,healthy
S002,38,M,patient,schizophrenia
S003,41,M,patient,schizophrenia
S004,29,F,control,healthy

See Subjects and cohorts.

Step 3 — Assign group labels

Navigate to Cohort analysis (/areas/neurology/studies/{id}/cohort). Assign a group label to each connectivity matrix dataset in the assignment table.

Step 4 — Batch graph metrics

Run Batch graph metrics to compute BCT measures across all matrix datasets. Results feed cohort metric comparison and provide per-subject metric tables.

Step 5 — Network-Based Statistics

Run NBS when at least two groups exist:

  • Permutation-based edge-level testing with family-wise error correction
  • Significant connected components highlighted on edge heatmaps
  • Component size and intensity statistics
  • Exportable NBS report artifact

Step 6 — Compare cohort metrics

Run Compare cohort metrics for group-level differences in global graph measures — clustering, efficiency, modularity, small-worldness — with effect sizes and statistical tests.

Step 7 — Interpret and review run history

Use Interpret on the cohort page after NBS or cohort metrics complete. Review the full run history for the study on the Runs page.

Expected outputs

  • Per-subject graph metrics table across the cohort
  • NBS significant components with edge statistics and heatmap data
  • Group comparison table with effect sizes for global metrics
  • AI narrative citing NBS and cohort metric statistics
  • Run history with parameters and pipeline version for reproducibility

Example insight

The schizophrenia group shows significantly reduced modularity in the default-mode network alongside elevated betweenness centrality in prefrontal hubs, consistent with dysconnection patterns reported in the literature.

Typical analyses

AnalysisGroupsQuestion
Disorder dysconnectivityPatient vs. controlWhich edges show NBS-significant differences?
Treatment responseResponder vs. non-responderDid connectivity normalize after intervention?
Cognitive correlationHigh vs. low score splitDo graph metrics track cognitive performance?
LongitudinalBaseline vs. follow-upHow do network properties change over time?

Related guides