Fig. 1: fMRI and tau-PET processing.

A Surface rendering of the 200 region of interest (ROI)-brain atlas, based on which we estimated (B) inter-regional functional connectivity-based distance based on 1000 participants of the human connectome project. Specifically, subject-specific connectivity matrices (i.e., Fisher-z transformed correlations of ROI-specific preprocessed fMRI timeseries) were averaged across subjects, thresholded at a density of 30% (i.e., only the strongest 30% of connections were retained) and transformed to connectivity-based distance (i.e., shorter distance = higher connectivity). C For each ROI, we determined the average connectivity-based distance to the remaining 199 ROIs as a measure of weighted degree (i.e., global connectivity), which was subsequently rescaled between −1 and 1 to determine a (D) continuous mapping of connectivity from hubs to non-hubs. The same brain atlas shown in A was applied to tau-PET standardized uptake value ratios (SUVRs), and E two-component Gaussian mixture modeling was applied to transform tau-PET SUVRs to tau positivities, i.e., tau-PET SUVRs that have been cleaned from the off-target binding curve. F Surface renderings of group-average tau positivities are shown for each diagnostic group of the ADNI and BioFINDER sample. G Subject-specific tau positivity values were subsequently multiplied with the (H) scaled hub map to determine (I) hub-weighted tau positivity values, giving tau in hub regions a positive weight and tau in non-hub regions a negative weight. J Hub-weighted tau positivities and K tau positivities were subsequently averaged in order to (L) compute the tau hub ratio, i.e., a single numeric index indicating whether an individual tau PET pattern was more pronounced in hub regions (i.e., more positive) or non-hub regions (i.e., more negative), while adjusting for global tau levels. The mathematical equation for determining the tau hub ratio for subject-level data is shown in M. Source data are provided as a Source Data file.