SA97COMPUTATIONAL PSYCHIATRY MEETS GENETICS: REINFORCEMENT LEARNING IN EARLY PSYCHOSIS AND IN RELATION TO CLINICAL AND MOLECULAR GENETIC RISK FOR SCHIZOPHRENIA

Abstract

Computational psychiatry offers opportunities for increased mechanistic understanding of mental disorder and to aid in the integration between psychiatry and neuroscience. By developing and testing mathematical models of behaviour, the computational psychiatry approach aims to provide increased understanding in how disordered neurobiology manifests in particular clinical phenotypes. A computational approach may further aid in measuring intermediate (latent) phenotypes that may be more readily linked to causes (whether genetic or environmental) than observed phenotypes. However, as yet, computational psychiatry researchers have, in the main, focused on pathophysiology rather than attempting to integrate aetiological factors with computational models of behaviour. Here, we take a computational 2 psychiatry approach to investigate learning in psychotic illness, and test whether modelled (latent) learning variables relate to molecular polygenic risk for schizophrenia. Reinforcement learning (learning from feedback) has been proposed to be of potential mechanisitc importance in underpinning positive symptoms and/or negative symptoms including anhedonia and avolition.