A Reinforcement-Learning Model of Active Avoidance in Wistar-Kyoto and Sprague Dawley Rats: Experimental Dataset with Model Code and Output
Year: 2020
Type: article
Abstract: The purpose of this experiment was to compare between-strain differences regarding avoidance behavior. Active avoidance data were collected from 40 Wistar-Kyoto (WKY) and 40 Sprague Dawley (SD) rats during a foot-shock experiment. The experiment consisted of 12 sessions each composed of 25 trials. Each trial consisted of a warning period, a possible shock period (if avoidance did not occur), and an intertrial interval (ITI) representing a safety period. During warning periods, rats had an option to press a lever in order to avoid future shock. However, if a lever press did not occur, then a shock period began during which the rats could press the lever to escape the foot-shock. After an escape or an avoidance, a 180-second safe period began (ITI). The maximum time of a warning period and a shock were both 72-seconds each if the rat failed to lever press during both periods. Every session also began with a 60-second habituation period during which the rat could familiarize itself with the experimental cage. Lever presses during ITI periods were classified as inter-trial responses (ITRs) whereas lever presses during the habituation periods were classified as anticipatory responses (ARs). Lever presses during shock periods were labeled as escapes “E” whereas lever presses during warning periods were labeled as avoidances “A”. Lever press data were discretized to 12-second periods. Frequency counts for lever presses during each 12-second timestep can be found in the raw data files (located in the “ratRL_datafiles_Model_InputOutput” folder). An example name of this file is “S09.csv” representing an SD rat or “W09.csv” representing a WKY rat. A reinforcement learning, actor-critic model was applied to this raw data to determine different learning parameters for each rat. The code for this model fitting can be found in the folder titled “ratRL_ModelFitting_Code”. Based on learning parameters determined from the reinforcement learning model code, a simulation of theoretical rats running through the experimental protocol was created. This simulation can be found in the “ratRL_Simulation_Code” folder. Inputs to this simulation are found in the “ratRL_codedTrials_Sim_InputOutput” folder. The summaries for each experimental rat’s performance during the foot-shock avoidance experiment are labeled “Sum” files (ie S09sum.csv) and the learning parameters for each rat are stored in the file called “parm_listfile.csv”. Output files are labeled “out” (ie S09out.csv) and represent the performance of a theoretical simulated rat. The differences in learning parameters between the two rat strains can provide insight into anxiety-disorders as the WKY rats have been used as an animal model for anxiety. Learning parameters can also be mapped to certain brain regions in order to explore possible neurological differences in the two strains in future experiments.
DOI: 10.17632/d6ybdxzkwz
Author: Palmieri, John
Language: English
Topic: Memory and Neural Mechanisms
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W6906344751