Superagers Research Study

This repository contains the analysis scripts from our study on superagers, which explores their structural connectivity, functional connectivity, and structure-function coupling.

Usage

This repository provides details on the analyses for transparency. The dataset is not included, though it can be requested with appropriate ethical approval.

License

The code is available under the MIT License, allowing others to reuse and adapt it with appropriate credit.

Table of Contents

• Folders
  • superager_classification
  • fsaverage_masks
  • fmri_analysis
  • structural_analysis
  • structure_function_coupling
  • elastic_net
  • analyses

Folders

superager_classification

• Purpose: Clean data from our cohort to merge the relevant files and classify participants as superagers or non-superagers.
• Scripts:
  • check_invalid_nps.R: Saves comments about participants' neuropsychological data, to be read manually and remove invalid data.
  • cleaning_bbhi_data.ipynb: Cleans BBHI data for merge.
  • cleaning_bbhi_senior_data.ipynb: Cleans BBHI senior data for merge.
  • superager_classification.ipynb: Classifies participants as superagers or non-superagers.

fsaverage_masks

• Purpose: Transform the fsaverage Schaefer-200 atlas and recon-all aseg parcellation to combined atlases in native DWI and native-T1 BOLD space.
• Scripts:
  • 01_fsaverage_to_t1.py: Transforms Schaefer atlas from fsaverage to each subject’s T1 space.
  • 02_subcortical_to_t1.py: Extracts subcortical regions from aseg.mgz files.
  • 03_combine_t1_atlases.py: Stacks Schaefer cortical and subcortical masks into a single subject-specific atlas in T1 space.
  • 04_t1_to_dwi_bold.py: Transforms the Schaefer/subcortical atlases from T1 space into native DWI and native-T1 BOLD space.
  • run_pipeline.py: Runs all scripts above, looping through each timepoint and cohort.

fmri_analysis

• Purpose: Scrub the preprocessed fMRI data, extract timeseries data, and compute functional connectivity correlations.
• Scripts:
  • 01_scrubbing_fMRI.py: Scrubs BOLD images based on a Framewise Displacement (FWD) threshold.
  • 02_extract_subjects_timeseries.py: Extracts timeseries data from BOLD images using Schaefer/subcortical atlases.
    • Uses functions in extract_timeseries.py.
  • 03_compute_subject_functional_connectivity.py: Processes timeseries data and computes Fisher z-transformation.
    • Uses functions in compute_functional_connectivity.py.
• Notes:
  • The scripts used to preprocesses fMRI data are available in another repository.

structural_analysis

• Purpose: Uses Multi-Shell Multi-Tissue Constrained Spherical Deconvolution (MSMT-CSD) to calculate white matter tracts and extract structural connectivity matrices.
• Scripts:
  • tractography_parallelized.py: Runs complete tractography pipeline. Performs rigid-body coregistration (structural to diffusion), tissue response function estimation, Fibre Orientation Distribution (FOD) estimation using MSMT-CSD, and Anatomically-Constrained Tractography (ACT) with SIFT2.
    • Uses spm_coregister_parcellation.m: For the rigid-body coregistration.
  • generate_structural_matrices.py: Computes structural connectivity matrices.
    • Uses functions in compute_functional_connectivity.py.
• Notes:
  • The scripts used in DWI preprocessing are available in another repository.

structure_function_coupling

• Purpose: Prepare and calculate structure-function coupling (SFC) metrics.
• Scripts:
  • 01_convert_to_individual_matrix.py: Converts functional and structural connectivity data into individual 214x214 matrices.
  • 02_structural_functional_coupling.py: Computes and visualizes SFC.
  • plot_group_connectivity_figure.py: Plots a single subject's structural connectivity matrix, functional connectivity matrix, and SFC vector.

elastic_net

• Purpose: Prepare and run logistic elastic net to classify participants as superagers or non-superagers.
• Scripts:
  • prep_data_for_en.py: Prepares data for elastic net models.
  • prep_weighted_global_roi_averages.py: Computes voxel-weighted global, sensory, heteromodal, and memory-relevant network averages for SFC.
  • log_en.py: Runs logistic elastic net classification predicting superager status using cross-validation and permutation testing.
  • en_fdr.py: Reads elastic net nohup log files, extracts model-level p-values, and reports FDR-adjusted p-values for each model.
  • plot_feature_importance.py: Plots the top 20 features from SFC elastic net models as bar charts. Includes a brain plot panel with cortical and subcortical regions.
  • make_supplementary_table.py: Generates supplementary table listing top 50 features from each SFC elastic net model.

analyses

• Purpose: Conduct all non-elastic net related analyses.
• Scripts:
  • analyses.R: Runs full results pipeline: mixed-effects models for episodic memory, superager status, and SFC with FDR correction.
  • results.Rmd: R Markdown document that sources analyses.R and auto-fills the results section with model outputs.
  • plot_sfc_difference.py: Averages superager vs. non-superager SFC difference maps and saves a cortical/subcortical figure.
  • plot_sfc_em_cortex.py: Combines the figure from plot_sfc_difference.py with forest plots of stats from results.html.