Work

Frozen in Fear: Machine Learning Unveils Sex Differences in Rats’ Freezing Postures and Defensive Reactions

Year: 2023
Type: dissertation
Abstract: Females are twice as likely as males to be diagnosed with anxiety-related disorders, yet they are underrepresented in preclinical research. We investigated fear responses in female and male Long Evans rats to understand the boundaries surrounding fear. Anxiety and fear, as motivational factors, prompt animals to engage in defense responses that maximize chances for survival. Predatory Imminence Continuum (PIC) theory provides a framework for understanding defensive behavior and organizing behaviors according to threat proximity: pre-encounter, post-encounter, and circa-strike, representing anxiety, fear, and panic states. However, a limited understanding remains of the specific thresholds where anxiety transforms into fear and vice versa. In Experiment 1, we induced increased anxiety and fear, exposing female and male rats to varying foot shock intensities as punishment for freezing behavior to investigate reactions to threats. We hypothesized that rats experiencing lower fear levels would learn the avoidance response. Our findings showed male rats punished for freezing reduced freezing behavior compared to females at the same low intensity. We interpreted these findings as females did not reduce freezing because they were positioned higher on the PIC, specifically in the post-encounter mode, where defensive freezing dominates. Experiment 2 aimed to establish the gradation of fear in the post-encounter mode and determine its boundaries. Traditional fear assessment methods oversimplify freezing behavior as binary. Differentiating qualitative aspects of freezing postures can offer a deeper insight into fear, even when freezing durations are similar. We employed markerless pose estimation and developed a custom unsupervised machine learning (UML) algorithm to analyze freezing postures. We hypothesized the postures in which females freeze indicate their fear level. Our UML grouped freezing postures into eight distinct clusters; two were enriched with females punished for freezing. Female enrichment was not due to sexual dimorphism or shock reactivity. Analysis revealed that the animal's orientational, postural, and positional features also influenced clustering. Furthermore, animals engaged in different behaviors before and after each freezing bout, providing further insights into their fear levels. By understanding fear and its impact on females, we aim to focus on this understudied population and contribute to improving anxiety-related disorder diagnostics and treatment.
Author: Smith-Vickery, Nancy Jo
Language: English
Open Access status: green
OpenAlex ID: W7047308301
Cites: 173
Cited by: 1
Related to: 9