Abstract
Drosophila carrying the Shudderer (Shu) allele of the voltage-gated Na [+] gene paralytic display spontaneous convulsions and immobilization phenotypes that are exacerbated by high temperature. To automate identification of these aberrant behaviors in Shu mutants, we trained a machine-learning classifier on a manually annotated dataset. The system reliably classified walking activity and immobilization periods, while uncoordinated movement events were detected with moderate sensitivity and high specificity. We then characterized the behavioral repertoire of Shu mutants and wild-type flies over a temperature-ramp protocol (20 - 40 °C) using the classifier. Our developments facilitate quantitative studies of environmental or genetic factors that alter behaviors characteristic of Drosophila models of neurological disease.
