Decoding the Structure-Activity Relationship of the Dopamine D3 Receptor-Selective Ligands Using Machine and Deep Learning Approaches

Dysfunctions of the dopamine D2 and D3 receptors (D2 and D3) are implicated in neuropsychiatric conditions such as Parkinson’s disease, schizophrenia, and substance use disorders. However, their high homology presents significant challenges in developing subtype-selective ligands, crucial for elucidating receptor-specific functions and developing targeted therapeutics. In this study, we constructed machine and deep learning-based quantitative structure-activity relationship (QSAR) models to predict compounds’ binding affinity at the D2 and D3, as well as their selectivity for D3. High-quality training data were queried from the ChEMBL database, followed by a systematic filtration and curation process to ensure data reliability. Using eXtreme Gradient Boosting, random forest, and deep neural network (DNN) algorithms, we developed QSAR models with strong predictive performance, with DNN-based models slightly but consistently outperforming the tree-based models. A novel hyperparameter optimization protocol was developed to enhance the efficiency of building DNN-based models, and integrating predictions from all algorithms into a consensus approach further improved accuracy and robustness. Notably, our selectivity models outperformed the affinity models, likely due to noise cancellation achieved by subtracting the binding affinities of the two receptors. The Shapley additive explanations analysis revealed key pharmacophoric and physicochemical features critical for receptor affinity and selectivity, while molecular docking of representative D3-selective compounds highlighted the structural basis of D3 selectivity. These findings provide a robust framework for modeling ligand-receptor interactions and advancing the rational design of selective therapeutics for D2 and D3.