Variability vs. phenotype: Multimodal analysis of Dravet syndrome brain organoids powered by deep learning

Abstract

Dravet syndrome (DS) is a developmental epileptic encephalopathy (DEE) driven by pathogenic variants in the SCN1A gene. Brain organoids (BOs) have emerged as reliable models for neurodevelopmental genetic disorders, reproducing human brain developmental milestones and rising as a promising drug testing tool. Here, we determined the underlying molecular DS pathophysiology affecting neuronal connectivity, revealing an early onset excitatory-inhibitory imbalance in maturing DS organoid circuitry. However, neuronal circuitry modeling in BOs remains hampered by the notorious inter- and intra-organoid variability. Thus, leveraging deep learning (DL), we developed ImPheNet, a predictive tool grounded in BO live imaging datasets, to overcome the limitations of the intrinsic BOs variability. ImPheNet accurately classified healthy and DS phenotypes at early onset stages, revealing differences between genotypes and upon antiseizure drug exposure. Altogether, our DL-predictive live imaging strategy, ImPheNet, emerges as a powerful tool to accelerate DEEs research and advance toward treatment discovery in a time- and cost-efficient manner.