gaozhy / DGM-Sim

DGM simulations and applications with real data

For questions about these data and analyses write to [e-mail].

How to cite this work

Schwab, S., Harbord, R., Zerbi, V., Elliott, L., Afyouni, S., Smith, J. Q., … Nichols, T. E. (2017). Directed functional connectivity using dynamic graphical models. bioRxiv. doi:10.1101/198887

Notebooks with results and figures

• DGM-Simulations
• DGM-RSN10
• DGM-Mouse

Reproducing full analysis

• Clone this repository.
• Install all packages suggested at the top of the notebook.
• Adjust the PATH_HOME and PATH variables at the top of the notebooks.
• You may (re)estimate all networks with DGM but the notebook will load the already computed networks.
• As I will not provide the HCP and mouse time series, so time-series related chunks need to be disabled for the Notebook to completely run.

Obtaining the data

Network data

• All network data (from simulations, human fMRI and mouse fMRI) can be produced with DGM or loaded from the RData containers in the results folder.

Raw time series

• Simulation time series are based in the generative model described in [1] and are in this repository except sim1 and sim22 which can be obtained from the FMRIB NetSim Website. The Notebook will download these automatically and will extract them.
• Human RSN time series can be obtained from the Human Connectome Project, Parcellation+Timeseries+Netmats (PTN) from here.
• Mouse time-series is not open and must be requested from [e-mail], but the network data is available.

References

1. Smith, S. M., Miller, K. L., Salimi-Khorshidi, G., Webster, M., Beckmann, C. F., Nichols, T. E., et al. (2011). Network modelling methods for FMRI. NeuroImage, 54(2), 875–91. doi:10.1016/j.neuroimage.2010.08.063.

Additonal Analyses

• DGM-Stepwise