A cross-cohort replicable and heritable latent dimension linking behaviour to multi-featured brain structure

Abstract

Identifying associations between interindividual variability in brain structure and behaviour requires large cohorts, multivariate methods, out-of-sample validation and, ideally, out-of-cohort replication. Moreover, the influence of nature vs nurture on brain-behaviour associations should be analysed. We analysed associations between brain structure (grey matter volume, cortical thickness, and surface area) and behaviour (spanning cognition, emotion, and alertness) using regularized canonical correlation analysis and a machine learning framework that tests the generalisability and stability of such associations. The replicability of brain-behaviour associations was assessed in two large, independent cohorts. The load of genetic factors on these associations was analysed with heritability and genetic correlation. We found one heritable and replicable latent dimension linking cognitive-control/executive-functions and positive affect to brain structural variability in areas typically associated with higher cognitive functions, and with areas typically associated with sensorimotor functions. These results revealed a major axis of interindividual behavioural variability linking to a whole-brain structural pattern.

Data-driven identification of a left anterior hippocampus’ morphological network associated with self-regulation

“…Examining the relationships between these three matrices requires a multivariate approach. Partial least squares correlation (PLS) methodology (Krishnan, Williams, McIntosh, & Abdi, 2011) and Canonical Correlation Analysis (CCA) techniques (Mihalik et al, 2022) represent advanced multivariate statistical tools tailored for exploring relationship between brain markers/features (such as regional grey matter volumes) and other features or measurements (such as behavioral outcomes). Such multivariate approaches can also be used to explore multivariate covariance between features in one part of the brain (e.g.…”

“…1). CCA identifies brain weights (u) and multi-block weights (v), which describe linear combinations of the variables in X and in Y, respectively (Krishnan et al, 2011;Mihalik et al, 2022). These weights can be interpreted as a quantification of how much each original variable contributes to the latent dimension (Krishnan et al, 2011;Mihalik et al, 2022).…”

“…CCA identifies brain weights (u) and multi-block weights (v), which describe linear combinations of the variables in X and in Y, respectively (Krishnan et al, 2011;Mihalik et al, 2022). These weights can be interpreted as a quantification of how much each original variable contributes to the latent dimension (Krishnan et al, 2011;Mihalik et al, 2022). This model selects the weights in order to maximize the canonical correlation, which corresponds to the correlation of the brain scores (Xu) with the multi-block scores (Yv) (Krishnan et al, 2011;Mihalik et al, 2022).…”

“…The data was adjusted for age, age-squared, and gender within that framework. Upon identifying a significant latent dimension, its variance was systematically removed from the data through a deflation process, as outlined by (Mihalik et al, 2022). Subsequent to this deflation step, we sought another latent dimension.…”

“…In order to capture the large spectrum of the behavioral phenotypes (i.e. a global behavioral pattern spanning cognition, emotion and socio-affective aspects) associated with brain structural features, such as here hippocampal subregional and brain grey matter volume, multivariate approaches are required (Genon et al, 2022a;Nicolaisen-Sobesky et al, 2022). Third, any type of association between brain markers and behavioral markers (be it bivariate or multivariate) found in a sample/subsample is highly likely to result in a replicability failure when investigated in another sample/subsample.…”