[EN] Atypical sensory processing is a common feature of autism, yet the neural computations that give rise to these differences, particularly in relation to biological sex and etiological origin, remain unclear. Here we examine predictive auditory processing at the single-neuron level in the inferior colliculus of two adult rat models of autism: a genetic model with a heterozygous Grin2b deletion (Grin2b + /-) and an environmental model based on prenatal valproic acid exposure. We recorded neuronal responses to an auditory oddball paradigm and a cascade control sequence across lemniscal and non-lemniscal IC divisions under high-intensity stimulation, allowing us to derive indices of repetition suppression, prediction error and neuronal mismatch. Using generalized linear mixed-effects models that accounted for animal identity, inferior colliculus division, sex, and rat model, followed by hierarchical group-level comparisons, we identified robust alterations in predictive processing in both autism-like models. These effects varied across inferior colliculus divisions and differed between sexes, revealing distinct phenotype-specific signatures. The results indicate that sex and etiology jointly modulate early auditory computations in autism. More broadly, our findings highlight the translational value of predictive coding frameworks and support the use of complementary animal models to capture neurobiological heterogeneity across the autism spectrum
