Feature selection leads to divergent neurobiological interpretations of brain-based machine learning biomarkers
Dustin Scheinost
Nature Human Behaviour, Published online: 15 April 2026; doi:10.1038/s41562-026-02447-y Across more than 12,000 MRI scans, the authors show that features discarded during feature selection can rival top-ranked ones in predicting behaviour yet imply different brain circuits, warning that focusing only on top features can mislead neurobiology.