Skip to main content
. 2021 Nov 8;11:571. doi: 10.1038/s41398-021-01704-0

Fig. 2. Correlations between actual GCA scores and GCA scores that are predicted based on brain connectivity.

Fig. 2

We applied multivariate predictive models to 5937 subjects at 19 sites to identify brain-wide connectivity patterns that are associated with general cognitive ability (GCA). (Upper Left Panel) In leave-one-site-out cross-validation, functional connectivity patterns associated with GCA generalized to 19 out of 19 held-out sites. (Upper Right Panel) The overall mean correlation between observed GCA scores and predicted GCA scores (predicted exclusively from brain connectivity patterns) was 0.42, pPERM < 0.0001 (observed correlation was higher than all 10,000 correlations in the permutation distribution). (Lower Panel) Scatter plots for the six largest held-out sites (blue, orange, green, purple, brown, and pink) show highly consistent performance at individual sites.