Workflow for the generation and evaluation of bio-realistic, conductance-based single-neuron models

Code base for all-active model generation, evaluation and analysis (main engine: bluepyopt).

• Method developed and present in Nandi et al. (2020): https://www.biorxiv.org/content/10.1101/2020.04.09.030239v1

• Main developers: Ani Nandi, Werner van Geit, Tom Chartrand, Anatoly Buchin and Costas A. Anastassiou

• Genetic algorithm in action: selection + evaluation + evolution

Dependencies

• Python 3.5+
• Neuron 7.5 (compiled with Python support)
• BluePyOpt
• eFEL
• AllenSDK
• NeuroM
• Pandas
• Seaborn

Installation

bash pip install git+https://github.com/AllenInstitute/All-active-Workflow # install directly from the repository or bash git clone https://github.com/AllenInstitute/All-active-Workflow # clone repository from github cd All-active-Workflow pip install -e . # install in editable mode

Quickstart: Launching optimization jobs

In a Unix environment : bash $ source activate conda-env # conda environment with all dependencies $ launch_optimjob --help # Look at the options $ launch_optimjob --input_json job_config.json

• Configurable staged optimization

  job_config.json

  ```json { "ctyconfig": { "cellid": "483101699" }, "jobconfig": { "highleveljobconfig": { "condaenv": "ateamopt", "axontype": "stubaxon", "datasource": "web", "ephysdir": "ephysdata", "nonstandardnwb": false, "featurestimtypes": [ "Long Square" ], "featurenamespath": "featuresetall.json", "compiledmodfilesdir": "x8664", "jobdir": "483101699benchmarktimeout" }, "stagejobconfig": [ { "stagename": "Stage0", "stagestimtypes": [ "Long Square" ], "stagefeatures": "featuresetstage0.json", "stageparameters": "paramboundsstage0.json", "filterrule": "filterfeatprotopassive", "offspringsize": 512, "maxngen": 50, "optimconfig":{ "nengines": 256, "nnodes": 16, "qos": "celltypes", "nprocs": 16, "errorstream": "job.err", "outputstream": "job.out", "jobmem": "100g", "jobtime": "5:00:00", "ipyparallel": true, "ipyparalleldb": "sqlitedb", "mainscript": "OptimMain.py" }, "analysisconfig":{ "mainscript": "analyzestagejob.py" }, "seed": [ 1 ] }, { "stagename": "Stage1", "stagestimtypes": [ "Long Square" ], "stagefeatures": "featuresetstage1.json", "stageparameters": "paramboundsstage1.json", "filterrule": "filterfeatprotopassive", "offspringsize": 512, "maxngen": 50, "optimconfig":{ "nengines": 256, "nnodes": 16, "qos": "celltypes", "nprocs": 16, "errorstream": "job.err", "outputstream": "job.out", "jobmem": "100g", "jobtime": "5:00:00", "ipyparallel": true, "ipyparalleldb": "sqlitedb", "mainscript": "OptimMain.py" }, "analysisconfig":{ "mainscript": "analyzestagejob.py" }, "seed": [ 1 ] }, { "stagename": "Stage2", "stagestimtypes": [ "Long Square" ], "stagefeatures": "featuresetstage2.json", "stageparameters": "paramboundsstage2mousespiny.json", "filterrule": "filterfeatprotoactive", "APinitiationzone": "axon", "offspringsize": 512, "cpbackupdir": "checkpointsbackup", "maxngen": 200, "optimconfig":{ "nengines": 256, "nnodes": 16, "qos": "celltypes", "nprocs": 16, "errorstream": "job.err", "outputstream": "job.out", "jobmem": "150g", "jobtime": "12:00:00", "ipyparallel": true, "ipyparalleldb": "sqlitedb", "mainscript": "OptimMain.py" }, "analysisconfig":{ "mainscript": "analyzestagejob.py", "ipyparallel": true, "ipyparalleldb": "nodb", "errorstream": "analysis.err", "outputstream": "analysis.out", "nengines": 40, "nnodes": 4, "nprocs": 10, "jobtime": "10:00:00", "jobmem": "100g", "qos": "celltypes" }, "seed": [ 1, 2, 3, 4 ], "runhofanalysis": true, "runpericomparison": false, "depolblockcheck": true, "addfikink": true, "calcmodelperf": true, "modelpostprocess": true, "calctimestatistics": true, "timeout": 300, "hocexport": true } ] } }

  ```

HPC usage support

Instructions are included to run these compute intensive jobs on 4 different architectures (Allen Institute HPC cluster, NERSC, BBP5 and AWS).

Level of support

We are planning on occasional updating this tool with no fixed schedule. Community involvement is encouraged through both issues and pull requests.