Goal-planning in Psychosis: a study across humans, mice and neural networks

People with a schizophrenia diagnosis (PScz) have enduring impairments in planning-to-goals, which profoundly affect functioning. However, underlying brain mechanisms are unclear. Our labs have developed behavioural paradigms and computational tools that allow directly inferring planning algorithms from neuronal activity in the prefrontal cortex and hippocampus. We will leverage these tools to precisely pinpoint the computational basis of planning-to-goal impairments in schizophrenia. We propose a cross-species, multi-scale programme. We will use the same goal-planning tasks and neural measures across human PScz, mouse models and artificial neural networks. In mice, single neuron resolution recordings and optogenetic manipulations will reveal the mechanistic minutiae of computational impairments, and their permanent rescue using known pharmacotherapy. Human functional neuroimaging will translate mechanisms to PScz. A connected language study of hundreds of PScz across four continents will probe the same computations using clinically-scalable behavioural measures, helping demonstrate cross-domain generalizability and catalyse mechanistically-informed patient stratification. Artificial neural network models will bring this all together, teaching us how circuit-level dysfunctions in PScz impact computations behind planning, using simulations and neuroimaging analysis. Together, these three strands will synergise to provide unprecedented computational insight into one of the most important symptoms and recovery targets in schizophrenia.