Poster C106, Sunday, March 25, 1:00-3:00 pm, Exhibit Hall C
Identifying inter-relations between genetic polymorphisms and reinforcement learning: multivariate insights from behavior and computational modeling
Carrisa Cocuzza1, Jim Cavanagh2, Michael Cole1, Travis Baker1; 1Rutgers University, 2University of New Mexico
Background: Successful application of reinforcement learning (RL) is critical for daily decision-making. A neurocomputational theory posits that an individual’s ability to learn from positive and negative reinforcement can be predicted by genetic factors related to the midbrain dopamine system. However, support for this claim remains highly controversial. The purpose of this study is to expand upon those findings and apply structural equational modeling to identify the inter-relationships between genetic factors related to striatal and prefrontal dopaminergic functioning and optimal RL in humans. Methods: Data were collected from undergraduate students in two studies and concatenated here to yield a total sample size of N=280, dramatically increasing statistical power. Single-nucleotide polymorphisms (SNPs) of interest include DRD2-957, DRD4-521, DARPP-32-rs907, and COMT, and participants’ trial-to-trial training choices during a probabilistic reinforcement learning task were modeled using an algorithm (Q-learning) adapted from machine learning, which calculates separate learning rates associated with positive and negative prediction errors. Results: We identified significant bivariate differences between DRD4-allele groups on positive learning rate, and the interaction between COMT and DRD4 allele pairs significantly discriminated between positive and negative learning-rate parameters. No differences were observed for striatal dopamine SNPs. Conclusion: These findings point to a critical role for prefrontal dopamine expression in RL, which has been typically described in terms of subcortical mechanisms. Moreover, our structural equation model provided a theoretical framework for bridging the gap between genes, reinforcement learning, and psychiatric conditions, highlighting new directions into individuated and nuanced clinical assessment.
Topic Area: THINKING: Decision making