README.md

Dissociation of reliability, heritability, and predictivity in coarse- and fine-scale functional connectomes during development

Erica L. Busch, 2023

This directory contains scripts to prepare functional connectivity matrices from ABCD rs-fMRI data, run and apply connectivity-based hyperalignment on those connectomes, and run downstream analyses as presented in our JNeurosci paper.

ABCD participants were selected for this project using the organize_subjects.py script. Connectomes were computed pre-hyperalignment using build_aa_connectomes.pyand post-hyperalignment using build_cha_connectomes.py. Analyses were run over pairwise similarity matrices of subjects' connectomes; these were computed using connectome_similarity_matrices.py.

In this paper, we look at three specific metrics:

1. Reliability of individual differences in RSFC, which is computed over connectomes computed on split-halves of the timeseries data (see idm_reliabiity.py) among unrelated subjects.
2. Heritability, which is computed over connectomes using a sample of monozygotic and dizygotic twins (see h2_multi_abcd_wrapper.m).

• The h2_multi metric was introduced in Anderson et al. (2021) and code was adapted from this repo.

3. Prediction of neurocognitive scores from RSFC, where we look at the degree to which variance in neurocognitive scores (general cognitive ability & learning/memory) can be predicted by individual differences in RSFC. The initial analysis is performed with idm_prediction.py. Permutation tests over those scores are run with idm_prediction_permutations.py, and an additional analysis controlling for head motion in the prediction is performed with idm_prediction_motion_control.py.

• The idm_pred pipeline was based off Feilong et al. (2021) and code was adapted from this repo.