Fig. 2: Anatomy and virtual lesioning of impulsivity and neuroticism networks.

i-ii, Positive and negative networks that predict (a) impulsivity and (b) neuroticism. A positive network denotes the connections for which increased functional connectivity predicts an increase of a given behavioral trait; a negative network denotes the connections for which decreased functional connectivity predicts an increase of a given behavioral trait. Impulsivity is mainly predicted by positive connections within the motor/sensory network, and negative connections between motor/sensory and other canonical networks. By contrast, the functional connections that predict neuroticism were much more distributed throughout canonical networks. iii, We ran a virtual lesioning analysis by removing functional connections overlapping with all but a single canonical network and reran CPM iteratively to compare the relative importance of the canonical networks in prediction accuracy for each trait. The functional connections overlapping with the motor/sensory network alone result in similar CPM prediction accuracy for impulsivity (relative to CPM using whole-brain connectivity); the functional connections overlapping with the frontoparietal and the subcortical networks alone result in similar CPM prediction accuracy for neuroticism (relative to CPM using whole-brain connectivity); and using the functional connections overlapping with other networks show a diminished—but only to a relatively small extent— prediction accuracy.