IMPROVING TRANSLATIONAL INSIGHT USING SEQUENTIAL SAMPLING MODELS IN DRUG CHOICE

The translation of pharmacological treatments for AUD remains challenging despite advances in neuroscientific and molecular sciences. Reverse-translation of human models of decision-making to formally characterise pre-clinical discrete choice procedures offers a solution. We sought to assess whether 1) evidence accumulation models can capture alcohol-related decision-making dynamics in mice 2) can explain canonical behavioural shifts following devaluation and 3) characterise decision-making changes following treatment with a novel GLP1-ELP-FGF21 agonist. In Experiment 1, animals responded on two levers for 15% v/v alcohol and grain rewards in a discrete trials procedure. This allowed us to measure the choice made and lever response time. In Experiment 2, a subset of these animals underwent devaluation using within-subjects sensory specific satiety. Finally, a subset of animals were treated with a GLP1-ELP-FGF21 (1000nmol/kg) agonist. Evidence accumulation models captured the relationship between behavioural allocation and response time distributions well and indicated that differences in latent parameter accumulation rate underlie individual choice preference. Consistently, these models showed reliably slower accumulation rates for devalued choices under sensory specific satiety. This suggests that accumulation rate likely indexes cognitive representations of reward value, as has been theorised in human decision-making tasks. Importantly, treatment with a GLP-ELP-FGF21 agonist specifically reduced alcohol choice and altered its temporal dynamics without impacting grain choice. Formal modelling revealed a specific attenuation of the accumulation rate for alcohol choices. These findings show that evidence accumulation models afford sensitivity to the decision-making architecture underlying mice choice which may be useful in translating pharmacological treatments for substance use.