Code repository for the manuscript, "Human cognition involves the dynamic integration of neural activity and neuromodulatory systems" by Shine, J.M. et al. in Nature Neuroscience (2019).
The study involves conducting a spatial principal component analysis (PCA) on multi-task data from the Human Connectome Project (HCP; http://www.humanconnectomeproject.org/) and then tracking the trajectories (i.e., eigenvalues or time series of PCs [tPCs]) of the eigenvectors over time to interrogate the low-dimensional signature of cognitive function in the human brain.
A brief overview of the analysis plan:
- Concatenate parcel-wise data from 100 subjects across 7 tasks
- Run 'pca.m' on the concatenated data in MATLAB a) top 5 eigenvectors ('eigenvec.m') and eigenvalues ('eigenval.m') are stored in the repository, along with the XYZ coordinates and network assignment of each of the cortical parcels (n = 333); b) collapse the data according to the phase of the first tPC in order to estimate the low-dimensional manifold ('make_manifold.m');
- Plot the eigenvalues of the eigenvectors and compare these to: a) the combined task regressor; b) neurosynth Topic Maps ('topic_maps.m'; http://neurosynth.org/analyses/topics/; https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5683652/); c) time-varying network topology (github.com/macshine/integration; https://www.ncbi.nlm.nih.gov/pubmed/27693256) d) neuromodulatory receptor maps (http://neurosynth.org/genes/).
- Compare the results to block-resampled null data ('block_resampling.m');
- Calculate network controllability measures (https://www.danisbassett.com/resources.html) and compare to top eigenvectors.
Please contact mac.shine@sydney.edu.au if you have any further questions.