AllenBO_vs_PID

1. Extract more spike rasters 

   1. On local comp, 

   2. use Jupyter Notebook 

   3. to open buildSpkRasters.ipynb, 

   4. update the session #, 

   5. run full notebook, 

   6. rsync files over to HPC

2. Fill in gaps in mTE analysis

   1. On HPC, 

   2. cd /N/project/memLab_DeepLabCut/AllenBO_vs_PID/slurmLaunchDirs/slurmFiles_fullSweep

   3. If just filling in holes in existing analysis 

      1. sbatch slurm_AllenSpkRast2mTE_v1.1.sh

   4. If computing mTE for new rasters

      1. emacs AllenSpkRast2mTE_v1.1.py

      2. update sessSearchString & epchSearchString as needed

      3. escape emacs (ctrl+x, ctrl+c)

      4. sbatch slurm_AllenSpkRast2mTE_v1.1.sh

3. See if networks can be collapsed based on new mTE_byTarget results

   1. on HPC

   2. cd /N/project/memLab_DeepLabCut/AllenBO_vs_PID/scripts

   3. python CombineTargetsIntoNetwork.py

4. 
   

• Documentation

  • Document network combination script on IDTxl page

• Pushing analysis forward (non-coding todos)

  • Move network files for pre v1.1 folder system

    • CA1

    • After moving CA1, delete old mTE_byTarget folder & binnedRasters folders

  • Launch mTE analysis for the next set of spike rasters

    • epch_all is priority

    • Secondary:

      • *movie*

      • *drifting* 

• Advancing the analyses

  • Replicate Synergy in Rich Club results

    • Tools:

      • Create CIJ  - Convert biTE2mTE_*/fullNet.p files into wtdAdjMat.csv adjacency matrices to use as CIJ input to RC scripts with python idtxl environment to run
        python ~/AllenBO_vs_PID/scripts/ExportNet_biTE2mTE_v1.0.py

      • Compute Rich Club Stats - Process wtdAdjMat.csv CIJ matrix to compute rich club stats in matlab with [RCcoeffs, RCpValues, RCzScores, netDirs] = run_computeRCstats;

      • 
        

    • TODOs: 

      • Test if synergy is higher in the rich club (see Fig 4 in Main results)

        • Write wrapper for synInRCAllThresh to read appropriate columns from triadPIDs.csv and the modified fullNet.csv file, collect the outputs and then save them.

    • Main results:

      • Fig1 - methods

      • Fig2 - triad TE (mvTE?) in Log_10 bits per bit ranges from -3 to 0.5 with peak around -1.

      • Fig3 - Rich clubs as significantly above 1 for log_10 richness parameter (r_k) of >-2. Most networks had significant rich clubs for thresholds in the to 10-50% range.

        • HOW TO: [RCcoeffs,RCpValues,RCzScore] = computeRCstats(CIJ) 

          • (located in /docs(1) / Sam_MATLAB / Projects / SynInRichClubs / CodeForPublishing / synergyInRichClubs)

          • Will compute rich club curve over threshold and compute stats based on random permutations of network.

      • Fig4 - ~85% of network synergy is in rich club for single threshold. Across thresholds, only at thresholds higher than 10% does the % network synergy dip below 80%.

        • HOW TO: [meanSyn,propSyn] = synInRCAllThresh(network, synVals) 

          • (located in /docs(1) / Sam_MATLAB / Projects / SynInRichClubs / CodeForPublishing / synergyInRichClubs)

          • %   network: weighted, directed network (adjacency matrix)

            • All required values, and some additional, contained in: sess-* / epch-all_* /  biTE2mTE_cmiEst-JidtD_seqPerm-500_mxShft-10 / fullNet.p

          • %   synVals: Nx4 matrix of synergy data where each row is a triad. Column 1  is the receiver node, column 2 is the first transmitter, column 3 is the second transmitter, and column 4 is the synergy value.

            • All required values, and some additional, contained in: sess-* / epch-all_* /  biTE2mTE_cmiEst-JidtD_seqPerm-500_mxShft-10 / PID_SydneyPID / triadPIDs.csv

          • 
            

      • Fig5 - Mean synergy in rich club is correlated with rich club coefficient.

      • Fig6 - More triad nodes in the rich club, the more synergy.

        • Not main result but the actual relationship is likely clearest for this case. The main result of more synergy in the RC is likely only viable because of nonuniform sampling across these conditions. However, given the low numbers of triads in the mvTE networks, it is possible we don't have the sampling required to make this happen. 

        • HOW TO: [meanSyn,propSyn] = synInRCsingleThresh(network,synVals,RCcoeffs,RCpValues,pValThresh)  

          • (located in /docs(1) / Sam_MATLAB / Projects / SynInRichClubs / CodeForPublishing / synergyInRichClubs)

      • Fig7 - Computation ratio is very stable over triads (~23%). 

      • Fig8 - Computation ratio is very slightly negatively related to RC membership

    • Replicate Synergy varies as function of sender mutual information

    • Replicate Synergy downstream of recurrent connections results

• Refining pipeline (coding todos to improve generality & robustness)

  • Convert scripts into functions with 'def' and 'if __name__ == '__main__':' section so they can play nicely with eachother

There are rich clubs in the networks

• Environment

  • These aren't allowed by UITS

    • Set up SSHFS - check instructions here

    • Set up password-less SSH? - also see instructions here

Processed so far:  [791319847]

Processed so far: