A bridge between physiological and perceptual views of autism by means of sampling-based Bayesian inference

RODRIGO ECHEVESTE; ENZO FERRANTE; DIEGO H MILONE; INÉS SAMENGO

Lisboa y online

Simposio; 2021 Champalimaud Research Symposium (CRS21); 2021

Fundación Champalimaud

Theories for autism spectrum disorder (ASD) have been formulated at different levels: rangingfrom physiological observations to perceptual and behavioural descriptions. Understanding thephysiological underpinnings of perceptual traits in ASD remains a significant challenge in thefield. Here we studied the link between Bayesian computations under weakened priors, andinhibitory dysfunctions, neural variability, and oscillations in ASD subjects. We worked with arecurrent neural-network model which had been optimized to perform sampling-based inferencein a visual setting. Here we took two parallel paths: In one of them, we modified the probabilisticgenerative model under which the stimuli are assumed to be generated in order to increase prioruncertainty. In the other, we weakened the inhibitory connections in the neural network to inducean inhibitory dysfunction. We found that both paths lead to consistent results in terms of therepresented posterior distributions, providing support for the view that both descriptions mightconstitute two sides of the same coin. The dynamical properties of the network allowed us toconnect the higher uncertainty in the estimation of stimuli that results from weakened priors andthe increased neural variability observed in ASD subjects. Moreover, the model capturesprevious experimental observations in the autistic population in terms of gamma oscillationpower and peak frequency. We believe these results highlight the potential for the use of theemerging body of function-optimized neural networks as models to establish mechanistic linksbetween neural activity and computations in the cortex that go beyond the study of neurotypicalperception.