Fig. 1

Workflow of the SuperAger identification study. This figure illustrates the workflow of the study aimed at identifying SuperAgers among elderly individuals. (a) SuperAger classification is performed using SNSB-II cognitive tests and blood biomarkers, compared to the cognitive performance of individuals in their 40s. (b) The machine learning pipeline incorporates feature selection (RFE, BORUTA), data augmentation using fine-tuned large language models (GReaT), and model interpretation with SHAP to reveal key biomarkers associated with cognitive resilience.