Brad E. Sheese, Ph.D.

Overview: Automated Behavioral Experiments with Zebrafish

I've had a long-standing interest in solving the problem of how we can provide undergraduates the opportunity to propose and conduct their own high-quality behavioral research at a small liberal arts school. Historically, the Psychology Department at Illinois Wesleyan University has had a commitment to providing students with hands-on laboratory experiences. I've been working to expand these experiences by developing systems that allow individual undergraduate students to conduct high-quality, low-cost research with sufficiently large samples to samples to reliably detect moderate to small effects. I have been using Zebrafish as a model organism since they are commonly used in neuroscience and psychopharmacological research and they are inexpensive to acquire and to maintain in large numbers in the lab.

The main product of this work, developed in collaboration with Mark Liffiton, has been ATLeS, a low-cost open-source research system that (nearly) completely automates studies of learning and conditioning with adult Zebrafish. We began work in March 2014 and completed testing of our basic system in May of 2018. Our system offers real-time tracking so it can dynamically respond to fish behavior. Our system also offers automated data visualization and statistical analyses. We have designed the system to be inexpensive (less than $100 per tracking box), modular (as many tracking boxes can be added to the system as your lab space allows), and accessible to anyone (hardware can be built with off the shelf components. software is free for anyone to use or modify). Undergraduates have run a number of pilot studies with the system as we have been working toward finalizing the system. Slides from a seminar describing the genesis of the project are available here.

Why Zebrafish are a Good Solution for Improving the Quality of Undergraduate Research

Behavioral scientists are often interested in studying phenomena that have fairly small effects on behavior. Unfortunately, the size of our samples are also often quite small. This is particularly true for undergraduates studying humans or rodents at liberal arts schools.

I suspect we do students a disservice when we have them conduct research that cannot reliably detect the effects they are interested in. I've had many conversations with students conducting their own research that revolve around what kind of statistical analysis they can use when they didn't get the sample they intended. It's not a good conversation to have with people who are just learning the ropes. I think we'd better off starting to train people how to conduct research by having them conduct simple research that is methodologically sound.

Zebrafish are commonly used in biomedical, pharmacological, and neuroscience research. The fish are inexpensive to acquire (30 cents a fish) and maintain, so they allow us the possibility of getting sufficiently large sample sizes to study small to medium effects. Many of the assays that have been developed to studying learning and conditioning and pharmacological responses in rodents can be adapted for use with Zebrafish. Many drugs can be introduced directly into the water with the fish and do not require injection. Consequently, Zebrafish are a great way to introduce undergraduates to behavioral research, either through studies of learning and conditioning or the study of the effects of drugs on behavior.

Why Automate Tracking and Coding?

There were a few goals in automating tracking of Zebrafish behavior:

• Historically, behavioral work with adult Zebrafish has often relied on labor-intensive human coding of videotaped behavior. Consequently, much of the published research on Zebrafish behavior relies on relatively small sample sizes. Automating both the experiment and the behavioral coding (and providing the solution to researchers for free) should facilitate much larger studies.
• Real-time tracking of the fish allows us to have contingent real-time responses to behavior. Real-time tracking and response allows us to use operant and classical conditioning to study a wide range of research questions.
• Automated coding and statistical analysis allows us to run complete experiments very quickly. This in turn allows us to iterate and improve. Instead of spending a whole academic year to collect, code, and analyze, we can spend a few days, consider what we've learned, tweak the experiment, run it again, and see what happens. This systems greatly reduces the laboratory feedback loop.