F. Connectome Predictive Modeling

The third question we try to answer in this work is whether we can predict different aspects of in-scanner thought patterns based on functional connectivity. To do this, we rely on Connectome Predictive Modeling.

The notebooks involved in this part of the analyses are:

This part of the analyses is performed via the following notebooks:

1. S16a_CPM_subject_aware.CreateSwarm: this notebook generates all the swarm jobs needed to attempt prediction of each of the 14 targets (e.g., Thought Pattern 1, Thought Pattern 2, Images, etc.) 100 times, and also for the 10,000 null simulations used to establish statistical significance.

2. S16b_GatherSwarmResults: this script is used to gather results from all the iterations, and save them together into a single pandas dataframe. Instructions on how to use this program are provided in S16a_CPM_subject_aware.CreateSwarm.

3. S17_CPM_View_PredictionResults: this notebook reads the results from all real and null iterations (as gathered with the script listed in 2), and generates the plots that show the accuracy for real iterations in relationship to the 10,000 null simulations.

4. S18_CPM_Dashboard: this notebook can be used to explore the CPM models associated with the different predictions. It will show the circos plots and a basic glass brain view. Final Glass brain views shown on the paper were generated via CONN using CONN_CPM_BothContrasts_onBrain.