Genetic associations between brain and cognitive flexibility: A Mendelian randomization study

Cognitive flexibility (CF) is an important psychological function, but research investigating brain characteristics related to CF at the whole-brain level is limited.

To examine whether genetically proxied brain imaging traits are associated with cognitive flexibility, using Mendelian randomization (MR) under standard MR assumptions.

We performed a two-sample MR analysis using publicly available GWAS summary statistics from UK Biobank and imaging genetics consortia to evaluate evidence of genetically predicted associations between brain imaging traits and Trail Making Test Part B (TMTB) performance. All participants were of European ancestry. Data extraction and analysis started from May to November 2024.

The genetic instruments for 704 characteristics related to brain region volumes; brain sulcal morphology (sulcus depth and sulcus opening distance (SOD)); brain white matter (WM) microstructure (axial diffusivity (AD), fractional anisotropy (FA); mean diffusivity (MD); mode of anisotropy (MO), radial diffusivity (RD)); cortical structure (surface average (SA); thickness average (TA)); functional networks (amplitude of low frequency fluctuations (ALFFs), functional connectivity (FC) and structural connectivity (SC)). GWAS sample sizes ranged approximately from 18,101 to 47,226, depending on the imaging phenotype.

Trail Making Test Part B performance in 93,024 European-ancestry participants, defined in the source GWAS as the residual of the log10-transformed alphanumeric trail time after regressing out the log10-transformed numeric path time, so that the measure indexes set-shifting net of basic visual-scanning and psychomotor speed; higher values indicate a longer completion time and therefore worse performance.

In univariable MR, 80 brain imaging traits showed nominal associations with TMT-B performance after excluding signals flagged by heterogeneity or pleiotropy tests. After within-domain multiple-testing correction, 10 associations remained significant (Bonferroni-adjusted P < 0.05); using FDR correction, 21 associations survived within-domain FDR q < 0.05 and 6 survived global FDR q < 0.05. Effect sizes were (β range: -0.077 [-0.114 to -0.039] to 0.203 [0.079 to 0.328] change in standardized TMT-B time per SD change in exposure). In multivariable MR, 14 traits showed independent associations with TMT-B performance; key traits associated with shorter TMT-B time included left banks of the superior temporal sulcus volume, left ventral diencephalon volume, left thalamus volume, right inferior parietal sulcus opening distance, insular surface area, temporal gyrus thickness, selected fronto-temporal functional connectivity, and selected dorsal attention/visual and control/amygdala structural connectivity. Traits associated with longer TMT-B time included left superior frontal sulcus opening distance, inferior frontal gyrus ALFF, selected fronto-temporal functional connectivity, and selected dorsal attention/salience and salience/caudate structural connectivity.

These MR results are broadly consistent with prior neurocognitive literature and highlight additional brain imaging phenotypes that warrant follow-up. Given the limited testability of key MR assumptions-particularly for brain imaging traits and TMT-B as a proxy outcome-no causal claim is made, our findings have not yet been independently replicated in an external cohort, and they should be interpreted as hypothesis-generating and require triangulation with complementary designs (e.g., longitudinal and experimental studies).