Deep Learning Analysis of Gut Nutrient Stimulation on Mice Pose

Deep Learning Analysis of Gut Nutrient Stimulation on Mice Pose

Clay Baker does remote research

Summer 2020 Remote Research

My independent remote project in the Bohórquez Laboratory, “Deep Learning Analysis of Gut Nutrient Stimulation on Mice Pose” was funded by the URS. Our lab discovered synaptic communication from gut sensory epithelial cells, known as neuropod cells, to the vagus nerve in response to nutrient stimuli. But, the behavioral response induced by different macronutrients on this synaptic gut-brain circuit is not yet understood. To explore these responses, Dr. Laura Rupprecht and Winston Liu conducted behavioral assays in which mice, chronically implanted with catheters, received macronutrient infusion directly to the gut. These assays were captured via a Point Grey Firefly S camera (funded by the URS) and securely shared via the DukeDataService. I then used DeepLabCut, an open source machine learning software package, to help quantify changes in mouse pose. DeepLabCut is a pose estimation software, which can take in a small amount of labeled video frames and accurately label other frames of similar species. For my project, DeepLabCut was used to consistently determine and compile the position of a given mouse’s nose, tail, paws, and ears frame by frame into compact datasets.  Ultimately, I used these labels to determine distinct changes in pose and quantify rapidly developed preferences for infusion directly the gut. This work pilots a new methodology to explore the implications of neuropod cell signaling within our lab, and it creates a baseline for further evaluation of distinct mice responses to various nutrients. The URS provided funding for high quality camera to capture these behavioral assays, allowing for me to continue my work remotely with the Bohórquez Laboratory this summer. - Clay Baker