How to use ASCENT

Running the ASCENT Pipeline

See Running the Pipeline for instructions on how to run ASCENT. The first three stages of ASCENT are invoked by the command python run pipeline. This performs the necessary operations to prepare simulations of nerve stimulation. The final stage of ASCENT is invoked by the command python submit.py, which runs the simulations. After the simulations are run, the Query Class can be used to obtain the resultant data, and the Plotter Module can be used to visualize the results.

../_images/pipeline_stages.png — The stages of the ASCENT pipeline.

Submitting NEURON jobs

We provide scripts for the user to submit NEURON jobs in src/neuron/. The submit.py script takes the input of the Run configuration and submits a NEURON call for each independent fiber within an n_sim/. These scripts are called using similar syntax as pipeline.py: "./submit.<ext> <run indices>," where <run indices> is a space-separated list of integers. Note that these submission scripts expect to be called from a directory with the structure generated by Simulation.export_nsims() at the location defined by "ASCENT_NSIM_EXPORT_PATH" in env.json.

Cluster submissions

When using a high-performance computing cluster running SLURM:

Set your default parameters, particularly your partition (can be found in "config/system/slurm_params.json"). The partition you set here (default is “common” ) will apply to all submit.py runs.
submit.py will run in cluster mode if it detects that "sbatch" is an available command. To override this behavior manually, use Command-Line Arguments.
Many parameters which submit.py sources from JSON configuration files can be overridden with command line arguments. For more information, see Command-Line Arguments.

Other Scripts

We provide scripts to help users efficiently manage data created by ASCENT. Run all of these scripts from the directory to which you installed ASCENT (i.e., "ASCENT_PROJECT_PATH" as defined in config/system/env.json). For specific controls associated with each script, see Command Line Arguments.

`scripts/import_n_sims.py`

To import NEURON simulation outputs (e.g., thresholds, recordings of V_m) from your ASCENT_NSIM_EXPORT_PATH (i.e., <"ASCENT_NSIM_EXPORT_PATH" as defined in env.json>/n_sims/<concatenated indices as sample_model_sim_nsim>/data/outputs/) into the native file structure (see ASCENT Data Hierarchy, samples/<sample_index>/models/<model_index>/sims/<sim_index>/n_sims/<n_sim_index>/data/outputs/) run this script from your "ASCENT_PROJECT_PATH".

python run import_n_sims <list of run indices>

The script will load each listed Run configuration file from config/user/runs/<run_index>.json to determine for which Sample, Model(s), and Sim(s) to import the NEURON simulation data. This script will check for any missing thresholds and skip that import if any are found. Override this by passing the flag --force. To delete n_sim folders from your output directory after importing, pass the flag --delete-nsims. For more information, see Command-Line Arguments.

`scripts/clean_samples.py`

If you would like to remove all contents for a single sample (i.e., samples/<sample_index>/) EXCEPT a list of files (e.g., sample.json, model.json) or EXCEPT a certain file format (e.g., all files ending .mph), use this script. Run this script from your "ASCENT_PROJECT_PATH". Files to keep are specified within the python script.

python run clean_samples <list of sample indices>

`scripts/tidy_samples.py`

If you would like to remove ONLY CERTAIN FILES for a single sample (i.e., samples/<sample_index>/), use this script. This script is useful for removing logs, runtimes, special, and _.bat or _.sh scripts. Run this script from your "ASCENT_PROJECT_PATH". Files to remove are specified within the python script.

python run tidy_samples <list of sample indices>

Data analysis tools

Python Query class

The general usage of Query is as follows:

In the context of a Python script, the user specifies the search criteria (think of these as “keywords” that filter your data) in the form of a JSON configuration file (see query_criteria.json in Query Parameters).
These search criteria are used to construct a Query object, and the search for matching Sample, Model, and Sim configurations is performed using the method run().
The search results are in the form of a hierarchy of Sample, Model, and Sim indices, which can be accessed using the summary() method.

Using this “summary” of results, the user is then able to use various convenience methods provided by the Query class to build paths to arbitrary points in the data file structure as well as load saved Python objects (e.g., Sample and Simulation class instances).

The Query class’s initializer takes one argument: a dictionary in the appropriate structure for query criteria (Query Parameters) or a string value containing the path (relative to the pipeline repository root) to the desired JSON configuration with the criteria. Put concisely, a user may filter results either manually by using known indices or automatically by using parameters as they would be found in the main configuration files. It is extremely important to note that the Query class must be initialized with the working directory set to the root of the pipeline repository (i.e., sys.path.append(ASCENT_PROJECT_PATH) in your script). Failure to set the working directory correctly will break the initialization step of the Query class.

After initialization, the search can be performed by calling Query’s run() method. This method recursively dives into the data file structure of the pipeline searching for configurations (i.e., Sample, Model, and/or Sim) that satisfy query_criteria.json. Once run() has been called, the results can be fetched using the summary() accessor method. In addition, the user may pass in a file path to excel_output() to generate an Excel sheet summarizing the Query results. Finally, use the threshold_data() method to return a DataFrame of thresholds with identifying information.

Query also has methods for accessing configurations and Python objects within the samples/ directory based on a list of Sample, Model, or Sim indices. The build_path() method returns the path of the configuration or object for the provided indices. Similarly, the get_config() and get_object() methods return the configuration dictionary or saved Python object (using the Pickle package), respectively, for a list of configuration indices. These tools allow for convenient looping through the data associated with search criteria.

Example uses of these Query convenience methods are included in examples/analysis/.

plot_sample.py
plot_fiberset.py
plot_waveform.py

Specialized ASCENT plots

See Plotter Module for more info; example uses of this module are provided in examples/analysis.

Video generation for NEURON state variables

In examples/analysis/ we provide a script, plot_video.py, that creates an animation of saved state variables as a function of space and time (e.g., transmembrane potentials, MRG gating parameters). The user can plot n_sim data saved in a data/output/ folder by referencing indices for Sample, Model, Sim, inner, fiber, and n_sim. The user may save the animation as either a *.mp4 or *.gif file to the data/output/ folder.

The plot_video.py script is useful for determining necessary simulation durations (e.g., the time required for an action potential to propagate the length of the fiber) to avoid running unnecessarily long simulations. Furthermore, the script is useful for observing onset response to kilohertz frequency block, which is important for determining the appropriate duration of time to allow for the fiber onset response to complete.

Users need to determine an appropriate number of points along the fiber to record state variables. Users have the option to either record state variables at all Nodes of Ranvier (myelinated fibers) or sections (unmyelinated fibers), or at discrete locations along the length of the fiber (Sim Parameters).