Lost balance: synergy-redundancy dysfunction in Alzheimer’s disease

<title>Abstract</title> Alzheimer’s disease (AD) is the most common neurodegenerative disorder, yet reliable early biomarkers are still lacking, which greatly hinders timely intervention for AD. This study employed an integrated information decomposition framework to systematically investigate the dynamic evolution of brain synergy and redundancy information in different AD stages and their associations with cognitive function, aiming to explore potential novel biomarkers for AD early diagnosis. The results show that, compared with the cognitively normal (CN) group, the cognitively abnormal (CA) group exhibited a distinct pattern of enhanced synergy and reduced redundancy. This pattern difference is primarily distributed within the default mode network (DMN), and is significant even in early-stages of AD, indicating that synergy-redundancy information is a promising biomarker for AD early diagnosis. Despite the inclusion of cognitively impaired individuals, functional decoding analyses still revealed that synergy mainly supported higher-level cognitive and social functions, whereas redundancy was more related to sensorimotor and basic cognitive functions. Additionally, compared to the CN group, the CA group exhibited stronger positive synergy correlations in brain regions related to working memory and cognitive control, possibly reflecting compensatory mechanisms, while negative synergy correlations were observed in pain-processing regions, potentially linked to impaired sensory perception. Stronger positive redundancy correlations in motor-related regions but negative correlations in reward-related regions, may suggest AD-related selective reorganization of information processing. Finally, a machine learning model constructed based on synergy–redundancy information demonstrated superior performance in AD stage classification, providing new insights for AD early diagnosis and further validating the potential of synergy-redundancy information as early biomarkers.