Ramon NogueiraJan 10, 2023

#NeuroPaperThread #NeuroNewPaper

1) Our article “The geometry of cortical representations of touch in rodents” with @chrisXrodgers Randy Bruno and @StefanoFusi is finally out! In brief, we found that whisker contacts in mice S1 are represented in approximately orthogonal subspaces https://www.nature.com/articles/s41593-022-01237-9 🧵​👇​

The geometry of cortical representations of touch in rodents - Nature Neuroscience

Mice were trained to discriminate objects using their whiskers. The geometry of the neural representations recorded in somatosensory cortex was disentangled with small non-linear perturbations, allowing for generalization and flexibility.

Nature
Show threadRamon NogueiraJan 10, 2023

@chrisXrodgers @StefanoFusi
#NeuroPaperThread #NeuroNewPaper

2) This deep dive into neural representations follows a previous study led by
@chrisXrodgers
at the Randy Bruno lab, where the task and neural recordings are described in depth, and very interesting results about task dependency are presented https://www.sciencedirect.com/science/article/pii/S089662732100372X

Show threadRamon NogueiraJan 10, 2023

@chrisXrodgers @StefanoFusi #NeuroPaperThread #NeuroNewPaper

3) Shortly, mice were trained to discriminate between concave and convex shapes using only their whiskers. They accumulated touches across time and whiskers to perform the discrimination task

Show threadRamon Nogueira

#NeuroPaperThread
#NeuroNewPaper
@chrisXrodgers @StefanoFusi

4) We monitored whisker movement and contacts and used linear and non-linear classifiers to decode the identity of the shape and the choice of the animal on a trial-by-trial basis. We found that linear integration is sufficient for both shape’s identity and the animals’ choice

Jan 10, 2023 at 10:35pmWeb
Show threadRamon NogueiraJan 10, 2023

@chrisXrodgers @StefanoFusi #NeuroPaperThread #NeuroNewPaper

5) We built a simulation of the task and changed the general parameters of the shapes. Linear integration was also sufficient for a variety of shapes, distances, angles, and sizes. Non-linear integration of touches was necessary for tasks with complex semantics of groups of shapes

Show threadRamon NogueiraJan 10, 2023

@chrisXrodgers @StefanoFusi #NeuroPaperThread #NeuroNewPaper

6) Populations of S1 neurons were fit with encoding models of varying degrees of flexibility. We found that S1 activity was best explained by models that allowed for non-linear mixed selectivity rather than pure or linear mixed selectivity

Show threadRamon NogueiraJan 10, 2023

@chrisXrodgers @StefanoFusi #NeuroPaperThread #NeuroNewPaper

7) We realized that instead of a random and disorganized mixed code, the geometry of representations in S1 was best characterized by a low-dimensional scaffold with small perturbations towards higher dimensions. An approximately disentangled representation in orthogonal subspaces

Show threadRamon NogueiraJan 10, 2023

@chrisXrodgers @StefanoFusi #NeuroPaperThread #NeuroNewPaper

8) This organization was not a simple consequence of somatotopy because all whiskers were represented in all columns (see also Rodgers et al 2021). This representational geometry can be useful for balancing generalization and flexible behavior

Show threadRamon NogueiraJan 10, 2023

@chrisXrodgers @StefanoFusi #NeuroPaperThread #NeuroNewPaper

9) Finally, we fit RNNs to perform an artificial equivalent of the task and found that the geometry of the representations is determined by the difficulty of the task, and RNNs trained on complex tasks can still form representations that can generalize to unseen conditions

Show threadRamon NogueiraJan 10, 2023

@chrisXrodgers @StefanoFusi #NeuroPaperThread #NeuroNewPaper

10) Thanks for reading until the end. Many more details in the main article and supplementary figures. Thanks for the support from our many funding agencies and to the
Center for Theoretical Neuroscience, and the Zuckerman Brain at Columbia University. Please reach out if you have any questions or suggestions!

Show threadRamon NogueiraJan 10, 2023

@chrisXrodgers @StefanoFusi #NeuroPaperThread #NeuroNewPaper

11) Also, I am on the job market this year, so please do not hesitate to reach out if you think I could be a good fit for a theoretical/computational faculty position in your department!