How Vaping Roaches Can Teach Data Science

We may have been the ones who set cockroaches loose in the classroom. Which makes us thrilled to see ever more creepy-crawlies being enlisted to help democratize neuroscience! A new Arizona State University-led study, released just last month, shows how simple behavioral roach experiments can pair with free, open-source machine learning to make high-level science accessible to anyone.

Our very own SpikerBox has long shown that you don’t need a $20,000 lab to record spikes. The new study addresses another problem that often crops up when it comes to machine learning: the lack of processing power on simple computers that most schools have to contend with. All you need for the experiment proposed in the study is the cloud-based SLEAP algorithm to track and record roach movement, and Google Colab to analyze the data. (Both are free to use and require zero coding experience.)

You also need a simple elongated plastic chamber that you can 3D print, a syringe to pump in some air or vape vapor (yes, really), and 18 roaches. This way, even the most underserved K-12 and undergraduate schools can teach both hands-on ethology and data science in Python without spending a dime.

Can Roaches Get Hooked on Nicotine?

It’s not (entirely) a myth: cockroaches can thrive in just about any kind of environment. But the study’s proposed experiment asks another, more nuanced question. Do they actually like everything you throw their way? Are they likely to get into substances like nicotine, or even get hooked on it?

The roaches were first recorded while roaming free, scurrying along the plastic chamber at will. Then came the testing phase. When the researchers pumped nicotine-free vape towards the end of the track, the roaches were rather reluctant to scuttle towards the source of the vape, much less stick close to it.

Does this mean our hairy-legged house companions would never make enthusiastic vapers? The answer is: it depends. Pump in some 24% nicotine vape, and the roach will be much less averse to it. Instead of loitering about the opposite end of the chamber, it now seeks to explore—almost as if the nicotine left it wanting more. Plain vape? Bad. Nicotine vape? Not so bad!

“The experiment tests whether a cockroach has a preference to the drug (like nicotine) and/or the actual vape,” says co-author, Dr. Ulises M. Ricoy. Machine learning then serves to explore aspects of addiction in cockroaches such as drug seeking to characterize approach and overall movement behavior, but also learning—such as conditioned place preference, he goes on to explain. In other words, the experiment can teach students about roach learning, memory, and their drive to explore their surroundings.

Across two 2-to-3-day workshops that the authors tested it in, the participants reported a 94% bump in their knowledge of behavior tracking via machine learning, and over half said they wished to go on using the software. “Through these workshops that we have offered and will be offering again soon, educators get trained in how to use machine learning even if they can’t code at all. For example, educators and even teams of faculty and students can learn how to create a ‘skeleton’ as a crucial first step towards using other model organisms,” Dr. Ricoy adds.

That’s right: if cockroaches are too icky for your taste, it’s completely fine. The experiment can be tweaked and replicated on other creatures: plants, worms, or even ourselves. “The model is created for only the cockroach in this manuscript, but a model can be created for any organism to measure any behavior, not just addiction. These behaviors can be as simple as walking/running or as complex as the interactions with vapors that were tested in the manuscript,” another co-author, Dr. Jessica L. Verpeut, tells Backyard Brains.

The study confirms what we’ve been saying for years: the key to affordable neuroscience education has been lurking under our kitchen sinks all along!