Adel Ardalan (Qian Lab)
Where's Clifford? Recurrent Neural Network Models of Working Memory for Multiple Items with Continuous Values*

Many psychophysical experiments, as well as daily living tasks, involve keeping multiple items in working memory. When the items take continuous values (e.g., orientation, contrast and length) that need to be used or reported later, they must be stored in continuous structures of appropriate dimensions. In this talk, I show how such a structure might be represented in neural circuits by training recurrent networks to report two previously flashed stimulus orientations. I then discuss how the input and recurrent units interact to create and maintain such a manifold by adopting specific connectivity patterns and how these patterns change over time to implement both persistent and dynamic coding schemes. Finally, I discuss how we reproduce some psychophysics results from prior work.

*In collaboration with Chris Cueva, Ning Qian and Misha Tsodyks



Stephanie Herrlinger (Losonczy & Gogos Labs)
Simultaneous capture of heterogeneous sub-type in vivo interneuron dynamics in CA1 in a mouse model for the 22q11.2 deletion syndrome**

Interneuron (IN) dysfunction has long been implicated in Schizophrenia (SCZ). In human patients, subsets of interneurons have been identified as fewer in number or altered in their activity and connectivity. In mice, INs modulate place cell activity and stability, with evidence suggesting their disruption in the 22q11.2 deletion syndrome, the most commonly known genetic risk factor for SCZ. Most IN subtypes in the CA1 have not been examined in this disease context, along with their potential contributions to observed microcircuit deficits. Therefore, differential simultaneous characterization of IN subtypes in the CA1 within this model has the potential to elucidate specific mechanisms of hippocampal dysfunction in SCZ. To address this need, we have multiplexed in vivo IN 2-photon imaging with a post hoc comprehensive molecular characterization of all IN subtypes imaged by utilizing STARMap (spatially-resolved transcript amplicon readout mapping) to resolve subtype-specific IN dynamics during learning and behavior.

** In collaboration with Bovey Rao, Dr. Abbas Rizvi, Lexi Peterson, and Dr. Erdem Varol