Vanderbilt BrainHack 2026 Project
BrainHack Project: Build an open-source pipeline that transforms batches of mouse behavioral videos into standardized behavioral metrics, publication-ready figures, and analysis-ready datasets.
The Behavioral Inventory of Mouse Affective Pain (BIOMAP) is a machine learning–based behavioral assay developed in the HARMONIC Laboratory at Vanderbilt University Medical Center that quantifies pain-related behavior in freely moving mice using computer vision, pose estimation, behavioral classification, and custom analysis.
BIOMAP currently combines DeepLabCut, SimBA, and custom Python scripts. While pose estimation and behavioral classification are automated, many downstream analyses—including data processing, composite score calculation, figure generation, and statistical preparation—still require manual effort.
This BrainHack project aims to build a reproducible, open-source pipeline that automates the entire workflow, allowing researchers to process batches of behavioral videos with minimal user interaction.
Behavioral phenotyping often requires researchers to move data between multiple software packages, perform manual calculations, and generate figures by hand.
Automating BIOMAP will:
- Improve reproducibility
- Reduce manual processing
- Standardize analyses
- Simplify onboarding
- Enable larger behavioral datasets
- Support collaborative software development
Although originally developed to study sound-evoked pain, the resulting software could support a broad range of behavioral neuroscience applications.
Today's workflow consists of several independent analysis stages.
Batch of Behavioral Videos
│
▼
Run trained DeepLabCut models
│
▼
Tracking CSV files
│
├────────────────────────────┐
│ │
▼ ▼
BIOMAP Analysis Script SimBA
│ │
▼ ▼
Facial & Body Metrics Complex Behaviors
│ │
└──────────────┬─────────────┘
▼
Manual Post-Processing
▼
Manual Figure Generation
▼
Manual Statistics
For additional details, see:
Our goal is for an entire experiment to be analyzed with a single command:
biomap analyze data/raw_videos/The pipeline would automatically:
- Load batches of behavioral videos
- Run trained DeepLabCut models
- Perform behavioral classification with SimBA
- Calculate normalized behavioral metrics and composite scores
- Apply post-processing and quality-control steps
- Generate publication-quality figures
- Export analysis-ready datasets
allowing researchers to focus on scientific interpretation rather than data processing.
Develop an automated workflow that processes batches of behavioral videos with minimal user interaction.
Potential tasks include:
- Automating DeepLabCut execution
- Running SimBA within the pipeline
- Organizing intermediate files
- Calculating normalized behavioral metrics
- Computing composite scores
- Generating standardized figures
- Exporting analysis-ready tables
Develop methods to distinguish:
- Pain-related immobility
- Sleep
- Rest
- Quiet wakefulness
Potential approaches include:
- Pose-estimation features
- Temporal modeling
- Machine learning
- Rule-based methods
- Additional behavioral features
A fully validated classifier is not expected during the BrainHack weekend.
Contributors are encouraged to work wherever their interests and expertise align.
| Track | Example Tasks | Helpful Skills |
|---|---|---|
| Pipeline Engineering | Connect workflow components | Python, software engineering |
| DeepLabCut Integration | Automate trained DLC models | DeepLabCut, computer vision |
| SimBA Integration | Automate behavioral classification | SimBA, behavioral analysis |
| BIOMAP Analysis | Automate behavioral analyses | Python, data analysis |
| Figure Generation | Create publication-quality figures | matplotlib, Plotly |
| Quality Control | Build validation and QC reports | Statistics, testing |
| Documentation | Improve tutorials and onboarding | GitHub, Markdown |
| Experimental Design | Automate metadata and experimental structure | Neuroscience, data science |
No prior experience with every component is expected.
We welcome contributors from neuroscience, computer science, engineering, data science, machine learning, software engineering, and scientific visualization.
The goal is not to complete the entire pipeline in one weekend.
Instead, we hope to:
- Automate key portions of the BIOMAP workflow
- Integrate existing software components
- Improve reproducibility and documentation
- Reduce manual analysis and figure generation
- Build reusable tools that support future development
If you're new to the project:
- Read the BIOMAP bioRxiv Preprint
- Review the Current Workflow
- Explore the Desired Workflow
- Install the required software using the Software Overview
- Browse the repository's Issues tab
- 🎉 Start hacking!
Contributors will be acknowledged appropriately in future software releases, conference presentations, preprints, and publications in accordance with standard scientific contribution and authorship practices.