Life in the Lab

Using a mechanism that is likely much more complicated that we thought! Check out the lab’s most recent paper, from the inimitable Giuliano Iurilli, who performed the first systematic in vivo recordings ever in a part of the brain called the posterolateral cortical amygdala (plCoA), which couples odors (like fox odor or the scent of peanut butter) to different innate behaviors like attraction and aversion. Based upon anatomic and functional data (including from Sosulski et al), we and others predicted that odor representations in the plCoA should be pretty simple – probably labeled lines that either identified particular odors (i.e., we’d find that neurons in plCoA responded only to single innately relevant odors, so some neurons would respond only to fox odor while other neurons would respond only to peanut butter) or categorized odors based upon their behavioral meaning (i.e., we’d find some plCoA neurons that respond to all odors that are appetitive and other that respond to all odors that are aversive). Instead what Giuliano found was that individual neurons in the plCoA do not encode information about odor identity or odor category. Ensembles of neurons in plCoA appear to collectively convey information about the identity of odor objects in the world, and the structure of these ensembles is similar regardless of the identity or behavioral meaning of the odor object being encoded – at least in psuedopopulations of neurons assembled from multiple animals. In other words, odor representations in plCoA, which is known to be involved in innate behaviors, look just like those in the piriform cortex, which is involved in odor learning. This surprise suggests a dual role for the plCoA in both innate and learned behaviors, and new models for how the olfactory system can flexibly generate behaviors to innately-relevant odor cues in the environment. These data don’t definitively rule out labeled lines, but strongly suggest that the plCoA is doing much more than just acting as a relay. Check out his provocative – and exciting – paper here! Yoram Ben Shaul also wrote a great preview of the paper here (for those who want a little more context), and the paper was recently covered by the Simons Foundation (the work was supported in part by the Simons Collaboration on the Global Brain) here. Congrats, Giuliano!

…but not here. The folks at BioMed Central did a round-up of recent advances in behavioral quantification for neuroscience, and featured work by Alex, Matt, Jeff, Ralph and the rest of Team Behavior in the lab. It also includes a really nice summary of different approaches taken to this general problem, and includes bits and pieces about work by many of our friends and colleagues, including Gordon Berman, Kristen Branson, Megan Carey, Eiman Azim and Adam Hantman, all of whom are making great strides in developing modern methods for behavioral characterization. Blog post can be found at this link: http://blogs.biomedcentral.com/on-biology/2017/01/31/secret-language-behaviour/?utm_content=buffer4bcbf&utm_medium=social&utm_source=twitter.com&utm_campaign=buffer

….its intimate relationship with evolution. In addition to being the largest gene family in mammals, the odorant receptor genes are some of the fastest evolving. This suggests that perception is being sculpted on rapid timescales to enable individual species to detect and respond to those specific scents that are most critical for their survival and reproduction. Dan Bear (along with Jean Marc Lassance and Hopi Hoekstra) recently wrote a great review on what we can learn about smell by looking at evolutionary dynamics in Current Biology, which can be found here. It is part of a great review issue on evolution and the nervous system – check it out!

and can be found here. Note that the very next day the same blog posted a lovely pic from our friend Rachel Greenberg.

Check out this terrific trailer for our upcoming NIPS abstract (with Alex Wiltschko, Matt Johnson, Ryan Adams and David Duvenaud) put together by David – it does a really nice job explaining the value of merging a model-based approach with neural nets. The combination allows clear articulation of model structure – and maintenance of semantic meaning – while simultaneously taking advantage of flexibly learned feature embeddings. We think this is going to be an important and general method for capturing structure in high-dimensional data. If you are at NIPS this year, check it out!