tank-lab-to-nwb

NWB conversion scripts and tutorials. A collaboration with the Tank lab, funded by the Simons Foundation.

Install

$ pip install tank-lab-to-nwb

Usage

There are two ways to go about converting Neuropixel and Virmen behavior data.

(1) The primary processing pipeline synchronizes the task data with the electrophysiology data through TTL pulse and writes the spiking output to the same NWB file.

The required arguments for the use of the relevant functions are denoted in the comments of their respective sections of the conversion script. These include the file or folder locations of the data to be converted to NWB format, as well as several optional fields such as Subject information (species/age/weight).

After editing the conversion script convert_towers_processed.py with the proper path for Neuropixel and behavior data, the conversion can be executed from the terminal:

$ cd tank-lab-to-nwb
$ python tank_lab_to_nwb/convert_towers_task/convert_towers_processed.py

Alternatively, the conversion can be done using a custom tailored jupyter notebook spikeinterface_pipeline.ipynb that can be launched from the terminal:

$ jupyter notebook notebooks/spikeinterface_pipeline.ipynb 

(2) The secondary processing pipeline does not synchronize with TTL nor writes the spiking output to the NWB file.

$ cd tank-lab-to-nwb
$ python tank_lab_to_nwb/convert_towers_task/convert_towers_raw.py

The NWBFile can be inspected by reading it from a python script:

from pynwb import NWBHDF5IO

file_path = 'TowersTask_stub.nwb'
io = NWBHDF5IO(file_path, 'r')
nwb = io.read()
print(nwb)

Alternatively, the NWB data can be visualized with nwb-jupyter-widgets in a jupyter notebook:

$ jupyter notebook notebooks/towers_task_custom_widget.ipynb

from pynwb import NWBHDF5IO
from nwbwidgets import nwb2widget
from tank_lab_to_nwb.nwbwidgets import custom_timeseries_widget_for_behavior
from nwbwidgets.view import default_neurodata_vis_spec
import pynwb


file_path = 'TowersTask_stub.nwb'
io = NWBHDF5IO(file_path, 'r')
nwb = io.read()

default_neurodata_vis_spec[pynwb.TimeSeries] = custom_timeseries_widget_for_behavior
nwb2widget(nwb)

Background

Behavioral data mapping

The behavioral data is contained in .matfiles similar to the form: PoissonBlocksReboot4_cohort4_Bezos3_E65_T_20180202.mat This matlab file contains a struct (log) which contains several fields relevant for conversion. The list of fields that are extracted from this struct can be found below.

NWBFile

Subject

LabMetaData

The lap specific metadata is populated in tank_lab_to_nwb/convert_towers_task/virmenbehaviordatainterface.py using the custom extension ndx-tank-metadata built for extending the NWB LabMetaData schema with the required fields:

• session_performance can be edited from virmenbehaviordatainterface.py.

Rig

Rig information is converted from a log.version.rig struct object to a dictionary like as in this example:

{'rig': 'NPX',
 'simulationMode': 1,
 'hasDAQ': 1,
 'hasSyncComm': 0,
 'minIterationDT': 0.01,
 'arduinoPort': 'COM18',
 'sensorDotsPerRev': array([2469.2, 2469.2, 2469.2, 2469.2]),
 'ballCircumference': 63.8,
 'toroidXFormP1': 0.3879,
 'toroidXFormP2': 0.392,
 'colorAdjustment': array([0. , 0.4, 0.5]),
 'soundAdjustment': 0.2,
 'nidaqDevice': 1,
 'nidaqPort': 1,
 'nidaqLines': array([ 0, 11], dtype=int32),
 'syncClockChannel': 5,
 'syncDataChannel': 6,
 'rewardChannel': 0,
 'rewardSize': 0.004,
 'rewardDuration': 0.05,
 'laserChannel': 1,
 'rightPuffChannel': 2,
 'leftPuffChannel': 3}

Mazes

Maze information is converted to a DynamicTable object that can be converted to a pandas dataframe by calling .to_dataframe() as in this example:

Epochs

Trials

Behavior

Position, ViewAngle, Velocity, Collision