Predicting lifespan of Drosophila melanogaster: A novel application of convolutional neural networks and zero‐inflated autoregressive conditional Poisson model

Abstract

A model to classify the lifespan of Drosophila, the fruit fly, into short-and long-lived categories based on a sleep characteristic, extracted from activity data, is developed using a two-stage process. Stage 1 models the per-minute activity counts of each fly using a zero-inflated autoregressive conditional Poisson model. These probabilities are allowed to vary hourly, reflecting the circadian and other cycles present in a fly's sleep architecture. A 5-day moving window is used to model data allowing the model parameters to vary over the course of the fly's life. The resulting probabilities capture information about changes in sleep patterns with age and are hypothesized to contain features that help categorize flies into short-and long-lived groups. The resulting hourly zero-inflation probabilities over a 24-day period are utilized to create a 'heat map' containing information on the 24-hour daily sleep cycle and its changes across the 24-day observation period. In Stage 2, the heat maps for individual flies are used as inputs to a convolutional neural network to build a classification model. The estimated model provides a reasonably accurate way to group flies into lifespan categories. Grouping flies into such categories would facilitate the discovery of biochemical markers of aging.