Read on Twitter
Our work on 'Distributed coding of choice, action, and engagement across the mouse brain' is published today in Nature! With @peterzh, @MatteoCarandini, and @kennethd_harris. Here's a quick recap and what's new since bioRxiv. https://www.nature.com/articles/s41586-019-1787-x
In this study we used Neuropixels to record from ~30000 neurons in more than 40 regions of the mouse brain, both cortical and sub-cortical. During the recordings, mice performed a visual discrimination task.
We observed non-specific coding of action initiation in individual neurons from nearly every region we recorded. Action encoding is global.
Coding of the specific action chosen, however, was much rarer, yet still distributed across key regions: frontal cortex, basal ganglia, and midbrain. Choice was encoded in unexpected areas incl Midbrain Reticular and Zona Incerta. Coding was distinct in midbrain versus forebrain.
Using a new extension of the classic 'jPSTH' analysis, using canonical correlations to look at timing differences between population activity across simultaneously recorded regions locked to task events, we see indistinguishable timing across these different choice regions.
We saw that task engagement (trial-to-trial and comparing task with 'passive replay') correlates with a global pattern of changes: suppressed activity in cortex and enhanced activity subcortically. This could not be explained by measures of arousal, reward, or overt movements.
These results delineate the coding of choice, action, and engagement across multiple brain systems, showing that none of these are as localized as we might wish they were! We have a lot more to learn in the future about the mechanisms coordinating these distributed computations!
The data area all fully shared publicly now ( https://figshare.com/articles/steinmetz/9598406), and see links therein for code too. Thanks to @KavliFoundation and the efforts of talented Steinmetz lab member @reneee_wang, you can also access them in NWB format ( https://figshare.com/articles/steinmetz/11274968) if you prefer.
Last but not least, many thanks to the #Neuropixels consortium, especially Tim Harris, and its funders: @wellcometrust, @AllenInstitute, @HHMIJanelia, Gatsby. Thanks to @EU_H2020 and @HFSP for postdoc funding, and thanks @UCLBrainScience for being a great place to do this work!
