CCN Seminar with Marcelo Magnasco (Rockefeller University)
Please join us for a CCN Seminar with Marcelo Magnasco, professor at Rockefeller University. Dr. Magnasco is a faculty member in the David Rockefeller Graduate Program, the Tri-Institutional M.D.-Ph.D. Program, and the Tri-Institutional Ph.D. Program in Computational Biology & Medicine.
To schedule a meeting with Marcelo, please reach out to Jessica Hauser at email@example.com
Title: Topographical representation of complex sounds in rodent auditory cortex.
Abstract: Understanding how our brains parse and identify complex natural sounds has long been a central goal of auditory science. To explore how the brain analyzes combinations of just a few tones, we expose awake mice to sounds consisting of single or multiple tones; we use ultrawidefield 2-photon calcium imaging of a 4.4 mm2 area of temporal cortex containing the auditory areas, to measure activity of several thousand auditory-responsive excitatory pyramidal neurons simultaneously. The responses to single tones recapitulate well-known tonotopic organization results. The responses to two-tone combinations, however, display combination-specific features or nonlinearities, which we isolate by projecting the activity vector of the response to tone combinations perpendicularly to the activity vectors of the individual tones. Here we show that the magnitude of the nonlinear component of the neural response is consistent across all stimulus pairs and animals, and that the neurons most involved in this nonlinear response are spatially organized, with neurons of the same nonlinearity type (synergistic vs. antagonistic) spatially clustering together. Finally, we show that erasure of the nonlinear component drastically impairs the decodability of the neural responses. Our results are the first explicit demonstration that specific types of computations may be topographically organized, bolstering the long-standing conjecture that such organization subsumes computational roles and is not merely a byproduct of developmental algorithms.