Resting-state cortical hubs in youth organize into four categories

DOI: 10.1016/j.celrep.2023.112521

Authors:

Damion V. Demeter*, Evan M. gordon, Tehila Nugiel, AnnaCarolina Garza, Tyler L. Larguinho, Jessica A. Church

• Correspondence: ddemeter@ucsd.edu

---

Scripts and other resources for replication of this work

Scripts:

• Cortical Hub Identification Script: Identify_Hubs.py
  • This script identifies hub parcels and associated files.
• Hub Density Map Creation Script: Hub_Denisty_Map.py
  • This script creates a hub density map across all participants.
  • As in Figure 1 (A)
• Hub Profiles Script: Create_Hub_Profiles.py
  • This script creates hub profiles and a hub profile correlation matrix that can be used to cluster hub profiles into categories.
  • We recommend using the Louvain algorithm and methods described in this paper to identify hub categories, but other clustering methods can be used with this output.
  • The optional correlation matrix plot isn't as useful until after hub category clustering, but can be used to make a matrix as in Figure S3 (B).

Python Requirements: these scripts are written in python 3.9.7.

• Python package requirements:

Brain Connectivity Toolbox for Python (bct)
NetworkX (networkx)
NiBabel (nibabel)
Numpy (numpy)
SciPy (scipy) 

Helper Files:

1. COMBINED333_LR_Distance_MULTIPLICATION_MASK.csv
   • Used for distance censoring of participant z-transformed matrix (See Figure 6A).
   • NOTE: Only valid for data in Conte69_fs_LR space. (If using a different surface space, a distance multiplication mask will need to be made from that surface with combined hemispheres)
2. Gordon333_TEMPLATE.pscalar.nii
   • Used as a template to save density map pscalar.nii data.
3. Parcels_LR.dlabel.nii
   • Used for numbering, location, etc of Gordon 333 parcels.

Other Requirements:

1. Timeseries should be fully processed, motion corrected, etc (appropriate steps for your chosen processing pipeline). This current script requires that timeseries are created from the Gordon 333 Parcel set and exported to a .txt file. (This script does NOT handle vertex-wise data)
   • .txt file format should be 333 rows by X time/TR of scan, exported to .txt file using the Gordon 333 Cortical Parcel set
   • .txt files can be created from dense timeseries files (.dtseries.nii) by using the "wb_command -cifti-parcellate" command from Connectome Workbench
2. A .txt file for your participant list is required that has (space separated) {participant ID} {path to timeseries.txt file}. (see example in this repository for format)
3. Infomap should be installed locally and able to be called from the command line. (Tested with infomap versions 1.9.0 & 2.6.0 - Other versions may need adjusted clu file editing.)

Basic Outputs:

1. Identify_Hubs.py

   • /final_avg_pc_percs/ - Average participation coefficient percentile for each cortical hub (In Gordon 333 set parcel number order)
   • /final_csv_outputs/ - All distance censored zmat files
   • /final_hub_indices/ - Parcel indices (Parcel #) for parcels identified as a hub
   • /final_hubs_dlabels/ - dlabel.nii files with shaded hub parcels, used for visualization

2. Hub_Density_Map.py

   • /_Gordon333_Hub_Counts.txt - Cumulative count of how many hubs were identified for each parcel
   • /_hubs_density_map.pscalar.nii - Hub density map for viewing in workbench view

3. Create_Hub_Profiles.py

   • /final_conn_profiles/ - Hub profiles, per subID, for all identified hubs. (Figure 6 (b))
   • /_Hub_Profile_Correlations.png - Hub profile correlation matrix plot (Not very useful until after profile clustering)

• Beyond this, see the -h (help) argument in the scripts for full details of required arguments.