Schedule of Events | Symposia

Flexibility in FMRI Brain Dynamics Predicted Behavioral Stability and Flexibility

Poster Session A - Saturday, March 29, 2025, 3:00 – 5:00 pm EDT, Back Bay Ballroom/Republic Ballroom

Jean Ye¹ (jean.ye@yale.edu), Saloni Mehta¹, Milana Khaitova¹, Jagriti Arora¹, Fuyuze Tokoglu¹, C. Alice Hahn¹, Cheryl Lacadie¹, Abigail S. Greene², R. Todd Constable¹, Dustin Scheinost¹; ¹Yale University, ²Brigham and Women's Hospital

Altered stability and flexibility appear transdiagnostically. How the brain dynamically supports these operations and how their disruptions are linked to psychopathology remain elusive. One possibility is that the ability to flexibly engage recurring brain states supports stability and flexibility. Here, we leveraged a multivariate computational framework and external validation to investigate whether flexibility in brain state engagement predicts individual differences in stability and flexibility in three transdiagnostic datasets. Resting-state data were collected in the Yale (N=237), Transdiagnostic Connectome Project (TCP; N=163), and Consortium for Neuropsychiatric Phenomics (CNP; N=225) datasets. Flexibility was assessed with the Behavioral Rating Inventory of Executive Function (shift; Yale) and the Rumination Response Scale (TCP). Stability was evaluated by response time during the incongruent condition of the color word task (Yale & CNP). To avoid circular analysis, we identified four recurring brain states in 390 Human Connectome Project participants. Non-negative least squares regression tracked their moment-to-moment engagement in all participants. Flexibility in brain state engagement was estimated by variability in engagement (SEV) over time. We studied if overall SEV correlated with behaviors within dataset before external validation. Lower SEV was linked to worse stability in patients (Yale: r=-0.18, p=0.03; CNP: r=-0.37, p<0.01) and more inflexibility in all participants (Yale: r=-0.15, p=0.02; TCP: r=-0.16, p=0.04). Regression models trained in one dataset successfully predicted stability in patients (Yale-to-CNP: r=0.33, p<0.01; CNP-to-Yale: r=0.20, p=0.02) and flexibility in all participants from another dataset (Yale-to-TCP: r=0.22, p<0.01; TCP-to-Yale: r=0.17, p<0.01). These results indicate flexibility in brain state engagement underpins stability and flexibility.

Topic Area: EXECUTIVE PROCESSES: Goal maintenance & switching