Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster F in Poster Session F: Thursday, August 6, 1:45 – 3:30 pm, Kimmel Center, Shorin & Rosenthal Rooms

The Geometry of Drifting Spatial Representations

Yile Wang¹, Kaushik Lakshminarasimhan¹; ¹University of Texas at Dallas

Presenter: Yile Wang

Neural representations of environmental variables are seldom stable, a phenomenon called representational drift. Drift is ubiquitous in the brain, and the pattern of representational drift places key constraints on the compensatory mechanisms needed to maintain stable behavior. Yet, whether the pattern of drift is similar or different across brain regions is not known. To address this, we characterized and compared the geometry of drift in spatial representation using longitudinal neural recordings in the posterior parietal cortex (PPC) and the hippocampus (HIP) of mice navigating in virtual reality. For each brain region, we used an optimal linear transformation that combined rotation and scaling to align the low-dimensional neural manifolds of spatial position from different pairs of days. Prior to this transformation, the degree of cross-day misalignment was not significantly different in the PPC and HIP suggesting that the magnitude of drift is similar in these regions. Interestingly, linear transformation substantially reduced the degree of misalignment in PPC but left significantly larger residual misalignment in the HIP. Notably, pure rotation or scaling transforms alone also did not reveal any regional differences. These findings suggest that whereas representational drift in the PPC is largely linear, drift in the hippocampus is not. Maintaining a stable readout might therefore require region-specific mechanisms tuned to properties of the drift in the local circuit, rather than universal Hebbian-like mechanisms proposed in previous works.

Topic Area: Methods, Tools, Theory & Neural Coding