Abstract
Analysis of freely moving Drosophila captures complex movement behaviors. However, previous experiments have been limited by the inability to distinguish between falls and downward jumps (downjumps). Here, individual flies moving freely in a culture vial were recorded using a single 58 fps video camera. Upward jumps were readily identified by positive movement in the vertical direction. Several statistical and machine learning methods were used to distinguish between falls and downjumps, including Principal Component Analysis (PCA), K-Means Clustering, Uniform Manifold Approximation and Projection (UMAP), Hierarchical Density-Based Spatial Clustering with Applications to Noise (HDBSCAN) and Shapley Additive Explanations (SHAP). Falls were abundant and characterized by an initial velocity consistent with simple acceleration due to gravity. Downjumps were more rare, and were characterized by a greater initial velocity, indicating active propulsion by the fly. Aged flies took longer to resume movement after a fall, suggesting possible negative effects of falls. Falls in young w[1118]-strain flies exhibited mid-event velocities that were lower than expected, indicating some compensatory behavior that was reduced in aged flies. These methods should facilitate future studies of the effects of aging and neurodegenerative disease models on locomotor behaviors and falls, including the testing of potential interventions.
