Sex-specific network architecture and behavioral modulation of brain aging revealed by connectome-based predictive modeling

Despite evidence for sex differences in age-related changes in functional connectivity (FC), extant age-prediction studies have relied on mixed-sex samples. We applied connectome-based predictive modeling (CPM) on whole-brain resting-state FCs (parcellated into 100/16-cortical/subcortical regions) from the Human Connectome Project-Aging (age 36–100 years) dataset, separately for males (M; n = 263) and females (F; n = 339). Model performance and cross-sex generalizability were evaluated using Pearson’s r and externally validated with the Aging Adult Brain Connectome dataset (age 36–89 years; M/F: n = 143/201). Predictive edges were mapped to functional networks to characterize sex-shared and different networks. Within sex-different networks, we tested whether cognitive/emotional factors moderated age-related FC changes. CPM robustly predicted age within sex (M/F: r = 0.66–0.68/0.58–0.60; all p ’s < 0.001), across sexes (M → F/F → M r = 0.66/0.60), and in validation dataset (M/F: r = 0.70/0.57). Default mode network (DMN) contributed strongly to age-prediction across sexes, but network composition differed, with affective-control networks more significantly involved in females and sensory-perceptual networks in males. DMN-limbic connectivity change with age was negatively moderated by diminished fluid cognition in males and positively by enhanced negative emotion in females. These findings reveal sex-shared and different network mechanisms of brain aging and identify DMN-limbic connectivity as a convergent, age-sensitive circuit, influenced by individual differences in cognition and emotion across sexes.