Fig. 4: Human mental processes are organized into independent modes of SRN modulation.
From: Serotonin regulation of behavior via large-scale neuromodulation of serotonin receptor networks
a, FSL DR approach to derive SRNs in the HCP dataset. b, Individual differences in SRNs were tested for covariation with behaviors via CCA. Permutation inference CCA characterized two statistically significant orthogonal modes of population variation. Each mode captures a relationship between SRN functional connectivity measures (‘imaging cross-loadings’) and behavioral measures (‘behavioral cross-loadings’). Modes 1 and 2 are color-coded in blue and red, respectively, throughout the main figure and the related figure (Extended Data Fig. 4). c, For each mode, imaging cross-loadings are represented as a set of brain regions showing a specific fingerprint of SRN involvement (radial plot, where each axis represents a receptor type). (For un-thresholded imaging cross-loadings, see Extended Data Fig. 4a–c.) d, Scatter plots of significant CCA modes: top: CCA mode 1, r = 0.42; FWE-corr P = 0.0008; bottom: CCA mode 2, r = 0.39; FWE-corr P = 0.0253. Statistical significance assessed via 10,000 block-aware permutations and FWE-corr for multiple comparisons. The scatter plots illustrate the relationship between SRNs and behavior. e, For each mode, behavioral cross-loadings are represented as axial plots illustrating fingerprints of cognitive (left) and affective (right) variables, where each axis represents a behavior score (see Extended Data Fig. 4d for fingerprints on the full set of behavioral variables). f, For each mode, subjects were ranked according to their CCA subject score (which captures how well individuals are represented by the mode phenotype) and divided into ‘high scoring’ and ‘low scoring’ groups. Behavioral scores were then calculated for these two groups separately, for measures of delay discounting, reward and punishment, psychiatric scores (according to the Diagnostic and Statistical Manual of Mental Disorders (DSM)) and affect, to test whether the modes of variation identified by the CCA captured relevant differences in behavior in these domains. On each box, the central mark indicates the median, and the bottom and top edges of the box indicate the 25th and 75th percentiles, respectively. Each data point represents an individual (n = 812 individuals). *Significant difference between high- versus low-scoring group (one-way analysis of variance; Bonferroni corrected across two CCA modes and multiple variables within the domain of interest). The results in panel f show that while mode 1 primarily captures variance in delay discounting (P < 10−4), antisocial problems (P < 10−4) and anger/aggression (P < 10−4), mode 2 captures variation in reward (P < 10−4) and punishment (P = 0.0071), depression (P < 10−4) and panic/sadness (P < 10−4). Both modes showed a significant effect on life satisfaction (both, P < 10−4). FC, functional connectivity; T-stat, T-statistic.
