In-silico cognition signals aging and cognitive decline

Abstract

Cognitive performance declines with aging, yet current assessment tools such as neuropsychological testing and task-based fMRI face limitations including variability, learning effects, and limited standardization. Here, we introduce in-silico cognition, a reservoir computing framework that models cognitive task performance using individual structural brain connectomes. We use this framework to simulate four tasks targeting memory, executive, and perceptual-motor domains, generating standardized and reproducible measures of cognitive function. Across lifespan and independent replication cohorts, in-silico cognition declined with age and showed significant associations with white-matter integrity, locus coeruleus integrity, resting-state activation, and clinical cognitive performance. Among the four measures, memory capacity emerged as a strong marker of aging, accurately predicting individual age and distinguishing young from older adults. These findings show that in-silico cognition is a scalable, biologically inspired, and clinically translatable marker of cognitive aging.