Every few seconds, your brain quietly decides whether to pick a fight or pick up a date, and it never bothers to send you the memo. Fruit flies run the same two-channel program - aggression and courtship - just with smaller brains and fewer dating apps. The trouble has always been counting it. Watching a male fly headbutt a rival or serenade a mate is one thing. Measuring thousands of those micro-dramas, frame by frame, with the precision of a referee who never blinks? That has historically required machines that cost more than a used car.
A team publishing in eLife decided that was ridiculous, and built something better out of medicine blister packs.
The problem tastes faintly of gatekeeping
Here is the thing about studying behavior: the brain only gives up its secrets when you can measure what it produces. Want to know which neuron makes a fly aggressive, or which gene dials up courtship? You first need to count lunges, wing threats, tussles, taps, and the optimistically extended "courtship song" with brutal consistency. Do it by hand and you will lose your mind around video number three.
The existing automated tools work, but they carry two undertones nobody loved. First, they lean on rule-based algorithms - rigid "if the fly does exactly X, log behavior Y" recipes. Real behavior is messier than that, full of half-lunges and ambiguous wing flutters, and rigid rules miss the subtle stuff. Second, the hardware ran expensive enough to lock out smaller labs, classrooms, and any researcher working far from a well-funded institution. The science had a velvet rope, and the bouncer was a price tag.
Enter DANCE, with notes of duct tape and brilliance
The researchers built DANCE - the Drosophila Aggression and Courtship Evaluator - and the whole thing has a deeply satisfying finish of "wait, that's it?"
The hardware is constructed from medicine blister packs (yes, the kind your allergy pills come in) and acrylic sheets. Recording happens on an ordinary smartphone. No specialized cameras, no five-figure rigs, no climate-controlled imaging suite. Just clever geometry and a phone propped at the right angle.
The clever part lives in the software. Instead of rigid rules, DANCE uses supervised machine learning - six behavioral classifiers trained the way you would teach an intern. You show the system examples of a lunge, a wing threat, a courtship attempt, and it learns the texture of each behavior, including the borderline cases that trip up rule-based methods. The result is a tool that reads fly drama with more nuance, picking up dynamic variations that the old rigid recipes flattened out.
Does the cheap version actually hold up?
This is where it could have gone wrong - the budget setup that works in theory and falls apart on the bench. It didn't. The team validated DANCE across genuinely demanding conditions and it kept pace with the expensive gear.
They tested flies raised in social isolation versus enrichment (lonely flies behave differently, as anyone who has worked from home for too long can confirm). They ran neuropeptide knockdowns, switching off specific signaling molecules to watch behavior shift. They even used optogenetics, silencing targeted neurons with light to see which circuits drive aggression or courtship. Across all of it, DANCE captured the behavioral differences - and in the subtle cases, its machine-learning classifiers outperformed the rule-based old guard.
That matters because it means the affordable tool isn't a compromise. It is, in places, the sharper instrument.
Why this is quietly a big deal
DANCE finishes on a long note of accessibility. Behavioral neuroscience has leaned heavily on Drosophila for a century precisely because the fly's genetics and neural circuits are so tractable - a living model for how brains turn signals into decisions. Making the measurement tools cheap and open-source means a lab in a resource-limited setting, a teaching classroom, or a researcher working near a fly's natural habitat can now run experiments that used to demand serious capital.
More hands counting more flies means more chances to map how genes and neurons produce the universal urges to fight and to woo - the same ancient circuits that, in far more complicated form, are still running in the wet machinery between your ears. Sometimes the path to understanding the brain runs straight through a blister pack and a phone camera.
References
-
Yadav RSP, Dey P, Ansari F, Kottat T, Vasam M, Prabhu PP, Ayyangar S, Bhaskar S S, Prabhu K, Ghosh M, Agrawal P. DANCE provides an open-source and low-cost approach to quantify aggression and courtship in Drosophila. eLife. 2025. DOI: 10.7554/eLife.105465 | PMCID: PMC12747525
-
Kabra M, Robie AA, Rivera-Alba M, Branson S, Branson K. JAABA: interactive machine learning for automatic annotation of animal behavior. Nature Methods. 2013. DOI: 10.1038/nmeth.2281
-
Tao L, Ozarkar S, Bhandawat V. Quantifying influence of human choice on the automated detection of Drosophila behavior by a supervised machine learning algorithm. PLOS ONE. 2020. DOI: 10.1371/journal.pone.0241696 | PMCID: PMC7743940
Disclaimer: The image accompanying this article is for illustrative purposes only and does not depict actual experimental results, data, or biological mechanisms.