hide
Free keywords:
-
Abstract:
Mixed selectivity in neural codes is well documented in multiple brain regions, with individual neurons exhibiting tuning to several variables that are explicit or implicit in behavior. While this mixed selectivity has been observed in multiple brain regions, the scope of such selectivity, and the variables selected for, have never been documented on the scale of the entire brain itself.We examine single neuron firing using neural activity recorded by the international brain lab (IBL) in its brain-wide map: 583 neuropixel penetrations covering 361 brain regions defined by the Allen atlas. The recordings were made in mice performing a task in which mice maximize rewards by exploiting a blockwise stimulus probability governing the appearance of stimuli. The task features auditory inputs, visual inputs, and a variety of behavioral signals which we can examine in the context of single-unit activity.We fit generalized linear models to express single-unit firing as a function of task and behavioral regressors. For each neuron, a model is fit which describes spike counts in bins as a function of stimulus, feedback, wheel speed (absolute value of velocity), block stimulus probability, and first movement onset. The resulting weights governing the predicted response of the model are compared against a statistical null distribution, and the per-region proportions of significantly modulated neurons are reported.Preliminary results show brain-wide sensitivity to wheel speed and reward, and to a lesser extent, the block probability of trial stimulus side. Notably there is very little sensitivity to directional wheel velocity (i.e., signed speed). Global sensitivity to reward delivery is more unexpected, and to our knowledge not previously observed in the literature.The broad sensitivity to block probability within the trial after stimulus is also a surprising result. Because we have overlapping regressors during the within-trial period for both stimulus side and movement direction, it seems unlikely that this result is simply attributable to correlations with those variables. In future work we aim to further investigate the effect of expectation and the prior on neural activity using behaviorally-informed estimates of the animal’s internal prior. We also aim to investigate the basis of widespread responses to reward, and whether those responses can be explained by motor activity like licking not included as model regressors.
