Data Mining Algorithms Predict DAT1 and COMT Dopamine Genotypes Based on Reinforcement Learning Task

Reinforcement learning implies learning from positive and negative feedback when an agent interacts with its environment. Studies mapped feedback-based teaching signals to how the human brain processes information using dopamine (among other neurotransmitters). Two main dopamine genes; DAT1 and COMT...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:2023 IEEE World AI IoT Congress (AIIoT) S. 0678 - 0684
Hauptverfasser: Natsheh, Ashar Y., Jayousi, Rashid, Herzallah, Mohammad M.
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 07.06.2023
Schlagworte:
Biology
data mining
dopamine
feedback-based learning
Machine learning
Machine learning algorithms
Negative feedback
Neurotransmitters
prediction
Prediction algorithms
Reinforcement learning
Training
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Reinforcement learning implies learning from positive and negative feedback when an agent interacts with its environment. Studies mapped feedback-based teaching signals to how the human brain processes information using dopamine (among other neurotransmitters). Two main dopamine genes; DAT1 and COMT, have a key role in regulating dopamine levels in the brain. Each gene has two well-studied variations that modulate reinforcement learning differently, thus creating four DAT1-COMT interaction patterns. Every human being carries one out of the four patterns. Extracting genotype variations via biological sampling is costly and time-consuming. Here, we introduce a machine learning alternative for effectively predicting the DAT1-COMT genotype variations by training classifiers on data from a reinforcement learning task, namely: (1) . We k-nearest neighbor, random forest, and neural classifiers on reinforcement learning task data from 146 subjects who also provided blood samples for genotyping. The results showed that random forest has the best performance in predicting individual gene variations, while neural networks showed the highest performance for predicting the DAT1-COMT combined genotypes compared to biological sampling. Our approach opens a new direction for using machine learning in the form of reinforcement learning tasks and classifiers to infer key biological information.
DOI:10.1109/AIIoT58121.2023.10174308