Homotopic functional connectivity analysis resting-state fMRI

This repository contains all code and intermediate data needed to replicate analyses in our preprint, 'Mapping functional, cytoarchitectonic, and transcriptomic underpinnings of homotopic connectivity'.

About the data

Resting-state fMRI data

Two distinct datasets were included in this project, both of which are openly available: (1) the Human Connectome Project (HCP) and (2) the UCLA Consortium for Neuropsychiatric Phenomics LA5c study (CNP). The HCP subset used in this analysis is comprised of $N=207$ unrelated participants ($N=83$ males, $28.7 \pm 3.7$ years of age) from the S1200 release, as curated and preprocessed by the ENIGMA Consortium. Structural and functional connectivity matrices from the HCP dataset were pre-processed and made available by the ENIGMA Consortium.

The second dataset in this analysis (UCLA CNP) includes N=252 participants ($N=116$ controls, $N=48$ schizophrenia, $N=49$ bipolar I disorder, and $N=39$ attention-deficit hyperactivity disorder). Resting-state fMRI data, along with structural MRI, were downloaded from openNeuro and preprocessed in previous work (see Aquino et al. (2020) and Bryant et al. (2024)). Group-averaged functional connectivity matrices from the UCLA CNP dataset were compiled in Bryant et al. (2024) and are included here in the data/ folder for reproducibility.

Brain maps of topographical hierarchy

We examined three indices of topographical hierarchy from the neuromaps library: structural, functional, and transcriptomic. All map information is provided in the neuromaps documentation.

• Structural: hcps1200-myelinmap-fsLR-32k
• Functional: margulies2016-fcgradient01-fsLR-32k
• Transcriptomic: abagen-genepc1-fsaverage-10k

Parcellated data (in the 68-region Desikan-Killiany atlas) are included in the data/ directory for reproducibility.

Brain maps of physical embedding

We examined two aspects of physical embedding in the cortex: structural connectivity and Euclidean distance between homotopic region-region pairs. For structural connectivity, we used the group-averaged structural connectome (estimated from diffusion-weighted MRI tractography) from the same $N=207$ HCP participants as the fMRI dataset included in the ENIGMA Toolbox.

To compute the physical distance between each homotopic region--region pair, we calculated the Euclidean distance between centroid vertices (on the pial surface) per region in the fsaverage template space from FreeSurfer.

Parcellated data (in the 68-region Desikan-Killiany atlas) are included in the data/ directory for reproducibility.

Brain maps of vascular innervation

We examined two aspects of vascular innervation in the cortex: arterial density and venous density. For both properties, we queried brain maps published in Bernier et al. (2018), where the details of imaging acquisition and preprocessing can be found. Voxelwise arterial and venous density maps were downloaded from the Braincharter vasculature repository. Parcellated data (in the 68-region Desikan-Killiany atlas) are included in the data/ directory for reproducibility.

Replicating our analyses and figures

All analyses and visualizations are fully reproducible using a series of five Jupyter notebooks contained in the data_visualization/ folder.

Figure 1: Regional homotopic connectivity variation aligns with macroscale gradients of the topographical hierarchy

Components of this figure can be recreated with the notebook 1_Pearson_HoFC_vs_hierarchy.ipynb.

Figure 2: The macroscale HoFC regional gradient is preserved across neuropsychiatric disorders, despite global decreases in inter-hemispheric coupling

Components of this figure can be recreated with the notebook 2_Pearson_HoFC_neuropsychiatric.ipynb.

Figure 3: Homotopic connectivity is linked to macroscale functional network topology within and across hemispheres

Components of this figure can be recreated with the notebook 3_Pearson_HoFC_vs_network_topology.ipynb.

Figure 4: Regional HoFC is not statistically associated with structural connectivity or Euclidean distance, but it is inversely associated with arterial density

Components of this figure can be recreated with the notebook 4_Pearson_HoFC_vs_structural_vascular.ipynb.