Visual experience shapes functional connectome gradients

Cemal Koba, Joan Falcó-Roget, Olivier Collignon, Katarzyna Rączy, Marina Bedny, Mengyu Tian, Marcin Szwed, Anna-Lena Stroh

Code and data for the analyses reported in the paper above.

Abstract

The human cortex is organized along continuous functional gradients that capture systematic transitions in functional connectivity across the brain. These gradients describe large-scale organizational principles, including hierarchical transitions from unimodal to transmodal regions. Here, we provide the first characterization of cortical gradients and their coupling to cortical morphology in congenitally blind (n = 31) and sighted (n = 52) adults to assess the relative contributions of intrinsic (genetic) and experiential factors to cortical gradient organization. Using resting-state fMRI, we compared functional connectome gradients and their association with cortical structure. Both groups exhibited similar principal gradients: unimodal to transmodal, somatosensory to visual, and frontoparietal segregation, demonstrating that the fundamental scaffold of cortical organization emerges largely independently of visual experience. However, blindness altered specific features of the functional connectome: the visual network was more segregated from the sensorimotor network and more integrated with transmodal and frontoparietal networks. Moreover, blind individuals showed reduced canonical hierarchical ordering within early visual areas, and weaker structure-function coupling. These findings suggest that the development of large-scale cortical gradients reflects a genetically guided scaffold that is subsequently refined by sensory experience, which shapes both hierarchical differentiation and its embedding within cortical structure.

Repository Structure

.
├── data_tidy/                    # Processed data used as direct input to the analyses
│   ├── gradients.RDS              # Region-wise gradient scores (main cohort: SC/CB)
│   ├── gradients_lb.RDS           # Gradient scores including the late-blind (LB) sample
│   ├── minimum_distances.RDS      # Minimum distances within early visual cortex
│   ├── morphometrics.rds          # Cortical thickness/area/curvature/depth/volume per region
│   ├── vertexwise_sfc.RDS         # Vertex-wise structure-function coupling data
│   └── vertexwise_loadings.csv    # PC loadings for the vertex-wise structural PCA
├── results/                      # Supplementary tables referenced in the manuscript
│   └── table_s01.csv ... table_s23.csv
└── scripts/
    ├── R/
    │   ├── plotting.R              # Shared ggplot2 helper functions used across all figures
    │   └── utils.R                 # Data loading/reshaping helper functions
    ├── analyses_part1.qmd          # Region-wise gradient comparisons (SC vs. CB) + community-based (Yeo network) profiles
    ├── analyses_part1_lb.qmd       # Same as part1, including the late-blind (LB) sensitivity analysis
    ├── analyses_part2.qmd          # Visual hierarchy analysis and minimum-distance comparisons within EVC
    ├── analyses_part3a.qmd         # Structural PCA (thickness/area/curvature/depth/volume) and ROI-level structure-function coupling
    ├── analyses_part3b.qmd         # Vertex-wise structure-function coupling analysis
    └── renv.lock                   # Pinned R package versions for reproducibility

Quick Setup

The analyses were run in R using renv to pin package versions.

# from within scripts/
install.packages("renv")
renv::restore()

This installs the exact package versions recorded in scripts/renv.lock.

Analysis Flow

1. analyses_part1.qmd / analyses_part1_lb.qmd — region-wise and community-based (Yeo network) comparisons of gradient scores between groups
2. analyses_part2.qmd — visual hierarchy and minimum-distance analyses within early visual cortex
3. analyses_part3a.qmd — structural PCA across morphometric measures and its relationship to the functional gradients (ROI-level)
4. analyses_part3b.qmd — structure-function coupling at the vertex level

Each notebook sources R/utils.R and R/plotting.R, and reads its inputs from ../data_tidy/ and writes supplementary tables to ../results/.