Online Program

Return to main conference page

All Times EDT

Friday, June 5
Practice and Applications
Practice and Applications 4
Fri, Jun 5, 3:30 PM - 5:05 PM
TBD
 

Predicting the Lifespan of Drosophila Melanogaster: A Novel Application of Convolutional Neural Networks and Zero-Inflated Autoregressive Conditional Poisson Models (308402)

Presentation

Gayla R. Olbricht, Missouri University of Science and Technology 
V A Samaranayake, Missouri University of Science and Technology 
Matthew S. Thimgan, Missouri University of Science and Technology 
*Yi Zhang, Missouri University of Science and Technology 

Keywords: Modeling life-span, Fruit Flies, Neural Networks, Count data processes

A model to predict the life-span of the fruit fly is developed, using a two-stage process. The first stage involves the modeling of the activity data for each fly using a zero-inflated ACP model. The zero-inflation probability was allowed to vary from hour-to-hour over the 24 hours of a day to reflect the circadian and other sleep cycles present in a fly’s sleep architecture. A moving window of five-days was used to model data from five days at a time, allowing the model parameters to vary over the course of the fly’s life. The resulting probabilities captures information about how sleep patterns change with aging and is hypothesized to contain features that will help categorize flies into short and long-lived groups. The resulting zero-inflation probabilities for each day, over a 24 day period were then utilized to create a “heat map” that contains information on the 24-hour sleep cycle for each day as well as how this changes across the 24-day observation period. At the second stage, the heat maps for each fly were used as input to a convolutional neural network to build a predictive model. Results show that the prediction model provides a reasonably accurate way to group flies into life-span categories. Grouping flies into such categories would facilitate the discovery of bio-chemical markers of aging by assaying groups of short and long lived flies.