RNN reconstruction of mouse latent neural dynamics

Neural activity in response to sensory stimuli leads to a reorganization of neuronal ensemble dynamics, which can be captured by lower-dimensional latent dynamics. Recurrent Neural Networks (RNN) have been proven successful at extracting such latent representations of neural dynamics. Here, we applied an RNN to the data available in the Steinmetz Neuropixels dataset, containing electrophysiological and behavioral data from mice engaged in a visual go/no go contrast detection task. First, we chose as our input (‘seed’) region a subset of spike rates recorded from mouse primary visual cortex (VISp). We tested how well the latent space of the VISp seed area predicts the neural responses recorded in different regions of interest (ROIs): i) within the visual cortex; ii) within non-visual regions. We found consistent prediction errors when comparing the ability of the network to reconstruct target signals between visual and non-visual areas. Our results suggest that this RNN approach can extract low-dimensional dynamics from one brain region which can give important insights for interpreting neural dynamics of other brain regions.