If you have any problems related to the accessibility of any content (or if you want to request that a specific publication be accessible), please contact us at firstname.lastname@example.org.
Developing Automated Procedure to Quantify Mouse Behavior
AltmetricsView Usage Statistics
Animals have evolved many different behavioral responses to cope with sensory information. Understanding how sensory information can affect animal behavior will provide new insights into the quantification of animal behavior. The primary objective was to develop a reliable and quantifiable method of scoring different behaviors based on differences in genotype, speed of the stimulus, and internal states of the animal. Specifically, approach and freezing behaviors were the primary focus across all studies. In the first study, three groups of mice were used to test the effect of an altered N-methyl-D-aspartate receptor (NMDAR). NMDARs have two major subunits: NR2A and NR2B. The study included an NR2A knockout mice group, an NR2B knockout mice group, and a wild type mice group. All groups were placed into an arena and were exposed to a moving object that moved at different speeds across a computer monitor. The results showed that NR2A knockout mice experienced a higher frequency of freezes than other mice groups. In a separate study, three groups of mice had different variations of the Isl2-EphA3 gene, affecting the axonal projections to the superior colliculus. The three groups included homozygotes, heterozygotes, and control group mice. The results showed that the homozygotes had a noticeably higher frequency of approaches while the heterozygotes experienced more head twitches during freezing behavior. These results were consistent with previous studies showing the homozygotes to be more impulsive and develop Attention-Deficit/Hyperactivity Disorder (ADHD). Due to the consistency of the scoring data in these studies compared to the behavioral output of animals in other studies, a behavioral quantification method was developed through machine learning algorithms, including DeepLabCut and neural networks