Statistical Analysis and Modeling of the Behavioral Response to Humidity and Olfactory Cues by the Vinegar Fly Drosophila melanogaster

Abstract: Animals are impacted by the humidity of their surrounding environment, it affects their ability of thermoregulation, water retention and their overall well-being. For small insects the effects of the surrounding humidity is even more important as they have a limited water storage which depletes faster in a drier environment. Some insects such as the vinegar fly Drosophila melanogaster have developed the ability to sense humidity (hygrosensation) and navigate using humidity cues. Much research has contributed to our understanding of hygrosensation but the process is still not fully understood. In this thesis a framework is developed which is able to detect and visualise results from behavioural assays. The framework consists of a statistical analysis of trajectory data and a Gaussian mixture Hidden Markov Model (HMM) which simulates fly locomotion. A behavioural assay is also conducted in this thesis, in which vinegar flies are subjected to variations in humidity and their resulting trajectories are measured. The developed framework is applied to the experiment data in order to investigate hygrosensation. However as the reaction from flies was insufficient in the experiments of this thesis, the HMM is also applied to an external data set from an assay investigating olfaction, where distinct responses to stimuli were found. Furthermore when separate HMMs were fitted to trajectory data from different responses, the resulting model fits showed notable differences. Trajectories simulated from each model also differed, where one model showed flies moving towards the stimuli source at an increased speed and another showed flies moving sporadically with a reduced speed. The developed framework could be applied for analyzing further experiments investigating hygrosensation and serve as a starting point for analyzing trajectory data in general.