Search Papers | Poster Sessions | All Posters
Poster A97 in Poster Session A - Tuesday, August 6, 2024, 4:15 – 6:15 pm, Johnson Ice Rink
A mechanistic model of genetic effects on excitation-inhibition balance in a RNN model of auditory processing
Rebeca Ianov Vitanov1 (), Kate Baker1, Jascha Achterberg1,2, Danyal Akarca1, Duncan Astle1; 1University of Cambridge, 2Intel Labs
Neural networks offer a powerful tool for testing mechanistic hypotheses about cognition. We explored the utility of this approach in a monogenic developmental disorder, by integrating a recurrent neural network (RNN) model and MEG task data from individuals with ZDHHC9-associated intellectual disability and age-matched controls. Given experimental evidence of ZDHHC9 implication in inhibitory synapse formation, we tested whether reducing inhibition levels in a RNN model of auditory processing trained on neurotypical evoked responses recapitulates case group neurophysiology. We show that stronger reductions in recurrent, inhibitory weights resulted in increased peak amplitude and peak latency of RNN prediction relative to the pre-perturbation predictions, similar to case-control empirical trends. In contrast, increasing network excitation via the excitatory weights failed to consistently recapitulate these trends. Together, these results suggest that reduced synaptic inhibition is a plausible mechanism by which loss of ZDHHC9 function alters cortical dynamics during sensory processing.
Keywords: recurrent MEG ZDHHC9 intellectual disability