Like many labs doing computational ethology, we are on the one hand racing ahead in terms of building methods to organize information about spontaneous behavior, but on the other hand lack tools (conceptual or practical) for deciding whether one way of describing behavior is better or worse than any other (and hence whether anything we do helps or hurts). Here we take a small step towards addressing this challenge by taking advantage (as one does) of drugs. We use neuro- and psychopharmacology to induce behavioral variability in a cohort of hundreds of mice, and then ask about the ability of various behavioral characterization methods (including Motion Sequencing) to tell drugs apart. The work demonstrates that MoSeq is great at this kind of task, reveals something about why, and leads to a proposition that behavioral syllables are druggable targets that might be useful for building therapeutics. Congrats to Alex Wiltschko, Tatsuya Tsukahara and everyone else who helped out with this lovely work.

Check it out here!