DeepEthogram, a machine learning pipeline for supervised behavior classification from raw pixels.

Videos of animal behavior are used to quantify researcher-defined behaviors of interest to study neural function, gene mutations, and pharmacological therapies. Behaviors of interest are often scored manually, which is time-consuming, limited to few behaviors, and variable across researchers. We created DeepEthogram: software that uses supervised machine learning to convert raw video pixels into an ethogram, the behaviors of interest present in each video frame. DeepEthogram is designed to be general-purpose and applicable across species, behaviors, and video-recording hardware. It uses convolutional neural networks to compute motion, extract features from motion and images, and classify features into behaviors. Behaviors are classified with above 90% accuracy on single frames in videos of mice and flies, matching expert-level human performance. DeepEthogram accurately predicts rare behaviors, requires little training data, and generalizes across subjects. A graphical interface allows beginning-to-end analysis without end-user programming. DeepEthogram's rapid, automatic, and reproducible labeling of researcher-defined behaviors of interest may accelerate and enhance supervised behavior analysis. Code is available at: https://github.com/jbohnslav/deepethogram.

Circuit and molecular architecture of a ventral hippocampal network.

The ventral hippocampus (vHPC) is a critical hub in networks that process emotional information. While recent studies have indicated that ventral CA1 (vCA1) projection neurons are functionally dissociable, the basic principles of how the inputs and outputs of vCA1 are organized remain unclear. Here, we used viral and sequencing approaches to define the logic of the extended vCA1 circuit. Using high-throughput sequencing of genetically barcoded neurons (MAPseq) to map the axonal projections of thousands of vCA1 neurons, we identify a population of neurons that simultaneously broadcast information to multiple areas known to regulate the stress axis and approach-avoidance behavior. Through molecular profiling and viral input-output tracing of vCA1 projection neurons, we show how neurons with distinct projection targets may differ in their inputs and transcriptional signatures. These studies reveal new organizational principles of vCA1 that may underlie its functional heterogeneity.