Fig. 1: Selective impact of local brain stimulation on the low-dimensional trajectories of brain activity across working memory loads and experimental sessions.
a Spatial maps for the top three PHATE dimensions during n-back working memory tasks (1-back = red, 2-back = blue; the size of the spheres represents the relative weight of PHATE values; see “Methods”). Network communities adopted from Gordon et al.15 (“Methods”). The majority of nodes load onto task-positive brain regions constituting the visual, cingulo-opercular, dorsal attention, and frontoparietal networks. b The state-space embedding of trajectories of brain activity across working memory loads during ‘correct’ responses across baseline, S1 (task-irrelevant region targeted by TMS, indicated as yellow in a), and iPS (task-relevant region targeted via TMS, indicated as green in a). The size of the nodes represent the relative weight of PHATE values. The color of the trajectories indicates the engagement of brain regions belonging to specific network communities. Here, asterisks indicate significant values (*pFDR < 0.001). c For each session, we calculated the difference between the within-session group trajectory and each participant’s trajectory (total summed Euclidean distance) during 1-back and 2-back trials. The relationship between these load-related trajectory distances revealed a load-induced scaling in the baseline and the S1 sessions, but not in the iPS session. Source data are provided as a Source data file.
