4. The models did not accurately predict cognitive outcomes for everyone. We examined whether any clinical/demographic factors were associated with tendency to be misclassified. We found parental socioeconomic status and duration of antipsychotic medication use to be associated with misprediction.

3. While important connectivity features were widespread, we identified several key networks that were essential for accurate predictions using virtual lesioning.

1. Using functional imaging data from the Human Connectome Project for Early Psychosis, we found that connectomic models can accurately predict broad cognitive functioning in individuals with early psychosis.

Boston stands up for science!
#StandUpForScience