45 AutoGraph Behavior: Automating Data Processing in Neurobiology Research

Faculty Mentor: Moriel Zelikowsky (Neurobiology, University of Utah)

As computing technology becomes more prevalent in scientific fields like neurobiology, this presents an opportunity to increase the efficiency of scientific processes in research. One area that could be streamlined and automated is data analytics. Traditionally, researchers spend many hours manually entering behavioral data. This process is not only time-consuming but can also lead to human error. Once data is collected, researchers need to test the statistical significance of their findings. However, labs can face significant challenges in automating this data processing when there is a lack of familiarity with computational methods.

While existing commercial software like SPSS [1], GraphPad Prism [2], and Microsoft Excel offer data analysis capabilities, they present several drawbacks. These programs either fall short of fulfilling the entire data analysis pipeline from raw data to visualization or still require manual data entry. Furthermore, many require paid subscriptions, which limits accessibility and the potential for broader scientific collaboration and improvement. Often, multiple computer applications are needed for the analysis and visualization of experiments, which can be unnecessarily complex and leave greater room for error. Here, I designed an application, AutoGraph Behavior [6] which takes CSV files of mouse behavior data (optimized for Observer [3] or EthoVision [4] scoring) and creates graphs and performs statistical analysis with a point-and-click user interface.

AutoGraph Behavior is a Python desktop application designed to automate the analysis of neural, behavioral, and vocal data in rodents, automatically generating statistics and graphical visualizations. The program is a comprehensive, open-source application tailored to the specific needs of neurobiology researchers in the lab. It has a user-friendly GUI to improve research efficiency and accuracy, allowing researchers without computational experience to plot graphs and statistics. To achieve this, color conventions and font changes were applied to the user interface to enhance its visuals and aesthetics. Program usability was ensured by strategically placing buttons, inputs, and outputs in sensible locations for smooth human-computer interaction. The application automatically reformats data into a structure better suited for testing as needed and converts time-based data into floating point datatypes so that statistical tests in Python can run. Missing values are preserved as missing, and the program notifies the user when missing data is encountered. The user is also given the option to save the cleaned spreadsheet locally for viewing and editing. The software is designed to process datasets produced by many animal behavior tracking programs used in the lab, such as The Observer XT [3], EthoVision XT [4], and DeepSqueak [5]. The tool processes tabular data, and its functionality supports the analysis of behavioral data across different contexts, including the ability to perform t-tests for two groups and conduct comparisons for multiple groups using ANOVAs. During development, Python methods for these t-test and ANOVA calculations were added, and appropriate user input locations were created in the interface, following good software practices such as separation of concerns. To ensure the software’s reliability and accuracy, new functionalities and edge cases were tested one at a time, and any code containing bugs or inefficiencies was refactored. To validate the ability of AutoGraph to generalize across experimenters and datasets, it was tested by other researchers in the lab and found to produce accurate results. Finally, clear instructions were provided for users and future developers, and changes were documented and updated on GitHub [6] regularly to improve code readability. This tool was designed with the potential to shift from a lab-specific tool into a widely used resource for many research labs.

Bibliography

[1] IBM. (2022). IBM SPSS Statistics (Version 29) [Computer software]. https://www.ibm.com/products/spss-statistics

[2] Graphstats Technologies. (2023). Graphpad Prism (Version 10.1.0) [Computer software]. https://www.graphstats.net/graphpad-prism

[3] Zimmerman, P. H., Bolhuis, L., Willemsen, A., Meyer, E. S., & Noldus, L. P. J. J. (2008). The Observer XT [Computer software]. https://pubmed.ncbi.nlm.nih.gov/19587185/

[4] Noldus, L. P. J. J., Spink, A. J., & Tegelenbosch, R. A. J. (2001). EthoVision [Computer software]. https://link.springer.com/article/10.3758/BF03195394

[5] Coffey, K. R., Marx, R. E., & Neumaier, J. F. (2019). DeepSqueak [Computer software]. https://github.com/DrCoffey/DeepSqueak

[6] Mo, A., Grammer, J., & Zelikowsky, M. (2025). AutoGraph Behavior [Computer software]. https://github.com/ZelikowskyLab/autograph_behavior