Abstract
In mental health research, it has proven difficult to find measures of brain function that provide reliable indicators of mental health and well-being, including susceptibility to mental health disorders. Recently, a family of data-driven analyses have provided such reliable measures when applied to large, population-level datasets. In the current pre-registered replication study, we show that the canonical correlation analysis (CCA) methods previously developed using resting-state magnetic resonance imaging functional connectivity and subject measures (SMs) of cognition and behaviour from healthy adults are also effective in measuring well-being (a ‘positive–negative axis') in an independent developmental dataset. Our replication was successful in two out of three of our pre-registered criteria, such that a primary CCA mode's weights displayed a significant positive relationship and explained a significant amount of variance in both functional connectivity and SMs. The only criterion that was not successful was that compared to other modes the magnitude of variance explained by the primary CCA mode was smaller than predicted, a result that could indicate a developmental trajectory of a primary mode. This replication establishes a signature neurotypical relationship between connectivity and phenotype, opening new avenues of research in neuroscience with clear clinical applications.
Highlights
IntroductionOriginal study by Smith et al (2015)
From the remaining 14 301 subject measures (SMs), we identified Adolescent Brain Cognitive Development (ABCD) SMs to include in our study via a one-toone matching of the remaining ABCD SMs with the 461 Human Connectome Project (HCP) SMs published by the original authors
Following the approach of Smith et al, we colour the points by an example SM; here we chose fluid cognition as measured by the NIH Toolbox as it was most similar to the fluid intelligence measure used in fig. 1b from Smith et al While fluid cognition does not appear to be associated with the relationship between SM and connectome weights in Mode 1, we observed a positive relationship in Mode 2 such that subjects with greater Mode 2 weights for both SMs and connectomes have higher fluid cognition scores
Summary
Original study by Smith et al (2015) In their landmark 2015 study, Smith et al [6] investigated the high-dimensional relationship between functional connectivity and behavioural and phenotypic measures using data on young adults aged 22–35 from the Human Connectome Project (HCP) [7]. To study the relationship between these subject measures (SMs) and functional connectomes, the authors calculated a 200-dimension group independent component analysis (ICA) functional parcellation for the 461 HCP subjects in their study. From this parcellation, they derived subject-level functional connectomes, which are symmetric matrices of the edge weights between all 200 nodes. Of the SM data available in the HCP dataset, Smith et al chose 158 measures that were quantitatively and qualitatively appropriate for analysis
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
