Mixed selectivity seen in OFC clusters that are defined by temporal response patterns

August 16, 2021

img_0561I recently read this preprint from Christine Constantinople’s lab. The paper is about encoding of decision variables in frontal cortex and I read it (and wrote this post) alongside Abby Muthusamy, a UCLA PhD student in my lab (left).

Approach: The authors used a value-based decision making task in which rats were simultaneously presented with randomized auditory clicks and visual flashes indicating reward probability and volume available at each side port. Upon training, rats reliably selected the side port with the greater subjective reward value; they’re good at it! The team performed k-means clustering and conditional clustering on single-unit recordings from the lOFC to interpret neural encoding of task variables (stimulus, reward attributes, reward outcome, reward history, and choice).

Take home message: Both clustering methods revealed 5 subpopulations of neurons (clusters) in the lOFC with distinct temporal profiles. Importantly, these subpopulations were not functionally distinct: all 5 clusters were modulated by all task variables. This was clear both from a rigorous analysis and also from a GLM, the results of which are quite clear and also beautifully presented (see Figure). Back to the main message: remember here that “mixed selectivity” refers to a property of single neurons. It means that they are modulated by more than one variable. Here, we are looking at clusters rather than single neurons, so must use the term loosely. Still, the overall finding appears very much in favor of mixed selectivity, in keeping with observations of mixed selectivity in non-human primates for economic decision variables in frontal cortex. The results of the current preprint potentially conflict with another paper in rats observing categorical encoding and NOT mixed selectivity, making me wonder what’s really going on in the frontal cortex!
Screen Shot 2021-08-16 at 11.12.23 AM
A final interesting point: one cluster was transiently “enriched” for signals relating to reward history, and these popped up just before the animal’s choice. This suggests that subpopulations in the lOFC not only mix value and stimulus signals, but also previous and current outcomes before choice, just when the rat needs them most.

Skeptic’s corner: The authors utilized two clustering methods, which strengthened their argument for 5 subpopulations arising in the lOFC during this task. Still, we would have liked to see a method that explicitly tested if clusters exist at all (such as PAIRS from Raposo et al. 2014, or the even better MCMC employed in this paper; it allows ones to identify both pure-selective and non-selective components). The downside with methods like k-means is that they find clusters whether they are there are not, rather than asking whether clusters are present vs. absent.

Also, we found it curious that  Cluster 3 (Fig. 1E, 2D) contained two seemingly distinct subpopulations of neurons, one peaking at trial start and the other peaking immediately prior to choice. We wonder whether the early peaking neurons within Cluster 3 should belong to Cluster 1 instead?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

anneslist

Highlighting female systems neuroscientists

Fairhall lab

Computational neuroscience at the University of Washington

Pillow Lab Blog

Neural Coding and Computation Lab @ Princeton University

Churchland lab

Perceptual decision-making and multisensory integration

%d bloggers like this: