Fig. 4

Correlation between global network measures and cognitive test scores. Plots showing the linear models between a global cognition (measured by mPACCtrailsB) and global efficiency at delay 4; b, c memory (measured by ADAS Q4) and global efficiency at delays 1 and 2 respectively and d clustering coefficient at delay 5; e global cognition (measured by ADAS13) and clustering coefficient at delay 2; f executive function (measured by Trail Making Test—Part B) and global efficiency at delay 3; g executive function (measured by Category Fluency (Animals)) and global efficiency at delay 1; h attention (measured by Trail Making Test—Part A) and clustering coefficient at delay 1. Each model includes cognition scores as the dependent variables, the network measures as predictors, while controlling for age, sex, education, and cognitive status. To account for the nonlinear relationships, the squares of the network measures were also included in the model. The gray areas show the 95% confidence intervals (CI) for the predictions and the dashed lines show the best fit. The dots correspond to show CN Aβ + (blue), MCI Aβ + (light green), and AD Aβ + (dark green) subjects. All results were corrected for multiple comparisons using FDR