Final Report Summary - DOPAMINE&PLASTICITY (Dopaminergic modulation of plasticity during social learning)
This project combines computational modelling, electrophysiology (EEG/MEG), functional magnetic resonance imaging (fMRI), genetics and neuropharmacology to investigate the modulation of synaptic plasticity by dopamine (DA) during learning about both social and reward-based information. The purpose of this project is to evaluate biologically and computationally interpretable models for obtaining quantitative indices of synaptic plasticity during learning using both electrophysiological and fMRI recording methods.
To pursue this goal, we developed a novel interactive game (cf. Figure 1) that combines both social and reward-based information to guide behaviour and probes how people learn about other agents’ “hidden” intentions and decide to follow or disregard their advice (Diaconescu et al., 2012a, 2013a).
Following behavioural assessment, this paradigm was used in an fMRI study of healthy volunteers. Behavioural indices of learning and fMRI recordings were obtained and analysed using state of the art modelling techniques, including hierarchical Bayesian learning models. Specifically, we tested whether the models could quantify trial-wise learning variables, such as social and reward-based prediction errors.
We showed that participants combine both social and non-social information when making decisions and predicting outcomes; however, the efficiency with which participants utilize the social cues (i.e. advice) predicts more accurate performance in this task. Finally, we demonstrated that social and reward-based prediction errors are encoded distinctly in dopaminergic and dopaminoceptive brain regions (Diaconescu et al., 2012b).
Furthermore, by combining genetic analysis with a model-based characterization of the neural mechanisms underlying prediction errors in a social context, we found that single nucleotide polymorphisms (SNPs), which influence DA metabolism play an important role in how people process social information (Diaconescu et al., 2013b).
In summary, we found that the Val158Met polymorphism of the COMT gene plays an important role in how participants process social PEs (Diaconescu et al., 2013c). This polymorphism results from a Valine (Val) to Methionine (Met) mutation, which affects the thermal stability of COMT, an enzyme responsible for DA degradation. Consequently, this mutation results in a 40% reduction of COMT activity (Meyer-Lindenberg et al., 2005).
Participants with the Met/Met polymorphism and therefore reduced activity of COMT and high concentrations of DA in the PFC showed reduced activity in the dorsomedial PFC in response to negative social PEs (Figure 2), whereas the processing of positive PEs induced stronger responses in the dorsolateral PFC (Figure 3).
Together with computational modelling studies that emphasize the role of DA in reinforcement learning and decision-making (Dayan, 2012; Montague et al., 2004), these results suggest that DA-dependent mechanisms similar to those active during reinforcement learning may also operate during social learning.
In the upcoming 3 months, we begin a joint pharmacological EEG and fMRI studies of healthy volunteers, employing a double-blind, placebo-controlled between-subject design with dopaminergic agents. We will specifically test whether the learning models that we have developed can quantify drug-induced changes in synaptic plasticity that occur as a function of trial-wise learning variables, such as social and reward-based prediction errors. Finally, we will use recent advances in model-based decoding to test whether our models can detect which pharmacological manipulation took place in any given subject. If these models allow for precise detection of functional DA receptor status, this research could pave the way for future development of non-invasive, model-based measures of neurotransmitter regulation of synaptic plasticity in individuals who exhibit deficits in social cognition, such as those diagnosed with autism spectrum disorder (ASD).
References:
Dayan, P. (2012). Twenty-Five Lessons from Computational Neuromodulation. Neuron 76, 240–256.
Diaconescu, A.O. Mathys, C., Weber, L.A.E. Daunizeau, J., Fehr, E., and Stephan, K.E. (2012a). Inferring on the intentions of others: Models of reciprocal learning during an interactive game. (Ascona, Switzerland: International Conference on Decision Making),.
Diaconescu, A.O. Mathys, C., Weber, L.A.E. and Stephan, K.E. (2012b). Inferring on the intentions of others: Model-based characterization of social information processing. (Beijing, China: Human Brain Mapping),.
Diaconescu, A.O. Mathys, C., Weber, L.A.E. Daunizeau, J., Fehr, E., and Stephan, K.E. (2013a). Inferring on the intentions of others: Models of reciprocal learning during an interactive game. Prep.
Diaconescu, A.O. Mathys, C., Weber, L.A.E. and Stephan, K.E. (2013b). Model-based neurogenetic characterization of social information processing in an interactive game. (Seattle, USA: Human Brain Mapping).
Diaconescu, A.O. Mathys, C., Weber, L.A.E. and Stephan, K.E. (2013c). Do we trust others’ advice? A model-based characterization of social inference in an interactive game. Prep.
Meyer-Lindenberg, A., Kohn, P.D. Kolachana, B., Kippenhan, S., McInerney-Leo, A., Nussbaum, R., Weinberger, D.R. and Berman, K.F. (2005). Midbrain dopamine and prefrontal function in humans: interaction and modulation by COMT genotype. Nat. Neurosci. 8, 594–596.
Montague, P.R. Hyman, S.E. and Cohen, J.D. (2004). Computational roles for dopamine in behavioural control. Nature 431, 760–767.
To pursue this goal, we developed a novel interactive game (cf. Figure 1) that combines both social and reward-based information to guide behaviour and probes how people learn about other agents’ “hidden” intentions and decide to follow or disregard their advice (Diaconescu et al., 2012a, 2013a).
Following behavioural assessment, this paradigm was used in an fMRI study of healthy volunteers. Behavioural indices of learning and fMRI recordings were obtained and analysed using state of the art modelling techniques, including hierarchical Bayesian learning models. Specifically, we tested whether the models could quantify trial-wise learning variables, such as social and reward-based prediction errors.
We showed that participants combine both social and non-social information when making decisions and predicting outcomes; however, the efficiency with which participants utilize the social cues (i.e. advice) predicts more accurate performance in this task. Finally, we demonstrated that social and reward-based prediction errors are encoded distinctly in dopaminergic and dopaminoceptive brain regions (Diaconescu et al., 2012b).
Furthermore, by combining genetic analysis with a model-based characterization of the neural mechanisms underlying prediction errors in a social context, we found that single nucleotide polymorphisms (SNPs), which influence DA metabolism play an important role in how people process social information (Diaconescu et al., 2013b).
In summary, we found that the Val158Met polymorphism of the COMT gene plays an important role in how participants process social PEs (Diaconescu et al., 2013c). This polymorphism results from a Valine (Val) to Methionine (Met) mutation, which affects the thermal stability of COMT, an enzyme responsible for DA degradation. Consequently, this mutation results in a 40% reduction of COMT activity (Meyer-Lindenberg et al., 2005).
Participants with the Met/Met polymorphism and therefore reduced activity of COMT and high concentrations of DA in the PFC showed reduced activity in the dorsomedial PFC in response to negative social PEs (Figure 2), whereas the processing of positive PEs induced stronger responses in the dorsolateral PFC (Figure 3).
Together with computational modelling studies that emphasize the role of DA in reinforcement learning and decision-making (Dayan, 2012; Montague et al., 2004), these results suggest that DA-dependent mechanisms similar to those active during reinforcement learning may also operate during social learning.
In the upcoming 3 months, we begin a joint pharmacological EEG and fMRI studies of healthy volunteers, employing a double-blind, placebo-controlled between-subject design with dopaminergic agents. We will specifically test whether the learning models that we have developed can quantify drug-induced changes in synaptic plasticity that occur as a function of trial-wise learning variables, such as social and reward-based prediction errors. Finally, we will use recent advances in model-based decoding to test whether our models can detect which pharmacological manipulation took place in any given subject. If these models allow for precise detection of functional DA receptor status, this research could pave the way for future development of non-invasive, model-based measures of neurotransmitter regulation of synaptic plasticity in individuals who exhibit deficits in social cognition, such as those diagnosed with autism spectrum disorder (ASD).
References:
Dayan, P. (2012). Twenty-Five Lessons from Computational Neuromodulation. Neuron 76, 240–256.
Diaconescu, A.O. Mathys, C., Weber, L.A.E. Daunizeau, J., Fehr, E., and Stephan, K.E. (2012a). Inferring on the intentions of others: Models of reciprocal learning during an interactive game. (Ascona, Switzerland: International Conference on Decision Making),.
Diaconescu, A.O. Mathys, C., Weber, L.A.E. and Stephan, K.E. (2012b). Inferring on the intentions of others: Model-based characterization of social information processing. (Beijing, China: Human Brain Mapping),.
Diaconescu, A.O. Mathys, C., Weber, L.A.E. Daunizeau, J., Fehr, E., and Stephan, K.E. (2013a). Inferring on the intentions of others: Models of reciprocal learning during an interactive game. Prep.
Diaconescu, A.O. Mathys, C., Weber, L.A.E. and Stephan, K.E. (2013b). Model-based neurogenetic characterization of social information processing in an interactive game. (Seattle, USA: Human Brain Mapping).
Diaconescu, A.O. Mathys, C., Weber, L.A.E. and Stephan, K.E. (2013c). Do we trust others’ advice? A model-based characterization of social inference in an interactive game. Prep.
Meyer-Lindenberg, A., Kohn, P.D. Kolachana, B., Kippenhan, S., McInerney-Leo, A., Nussbaum, R., Weinberger, D.R. and Berman, K.F. (2005). Midbrain dopamine and prefrontal function in humans: interaction and modulation by COMT genotype. Nat. Neurosci. 8, 594–596.
Montague, P.R. Hyman, S.E. and Cohen, J.D. (2004). Computational roles for dopamine in behavioural control. Nature 431, 760–767.
