Neuronal Interactions

Sebastian Seung, David Tank, Jonathan Pillow, 
Adrian Wanner, Mark Ioffe, Alexander Riordan


Preliminary data shows choice­-specific activity sequences in posterior cortical neurons in an evidence accumulation task. In order to get a mechanistic understanding of the neuronal network architecture and computation underlying the sequential activity we will compare neural coding properties with anatomical connectivity by combining calcium imaging with serial­-section electron microscopy. Towards that goal we will benefit from the Seung laboratory’s dramatic recent progress in automated neuron reconstruction by artificial intelligence, which will make it feasible to find connectomes of millimeter­scale cortical volumes. In addition, we will compare neural coding with physiological connectivity by combining calcium imaging with optogenetic inactivation of various brain areas and optogenetic perturbation at cellular resolution. This aim will leverage the Tank laboratory’s recent work on all-­optical assays of connectivity.

Much of this project focuses on retrosplenial cortex, which is of special interest because choice-specific activity sequences are more linear in this part of cortex than in other cortical areas. Furthermore, temporally specific inactivation of retrosplenial cortex causes behavior that mimics the response to deletion of incoming evidence. We will test two main prediction:

1. That the local circuitry underlying activity sequences will exhibit sequential connectivity, that is, that connections from earlier to later neurons will be stronger or more frequent than those from later to earlier neurons. 

2. That inactivation of the anterior M2 region of cortex will lead to a reduction in non­specific excitation in posterior cortices and alter sequence timing. 

Importantly, these example predictions are not intended to be comprehensive. Instead the generality of these techniques will enable us to test a wide variety of predictions that emerge from the neural circuit models that will be generated and refined based on data from these experiments and other components of the project.