Quantum Machine Learning Predicting ADME-Tox Properties in Drug Discovery.

Area of Science:

• Computational Chemistry and Cheminformatics
• Artificial Intelligence in Drug Discovery
• Quantum Machine Learning Applications

Background:

• Evaluating absorption, distribution, metabolism, excretion (ADME), and toxicity is crucial but time-consuming and expensive in pharmaceutical R&D.
• Artificial intelligence, big data, and cloud technologies are emerging as powerful tools for predicting molecular ADME-Tox properties.
• Quantum computing, combined with machine learning, offers potential advancements in high-throughput screening and cost reduction for drug discovery.

Purpose of the Study:

• To propose and validate a novel quantum machine learning (QML) framework for predicting ADME-Tox properties of chemical entities.
• To leverage quantum kernel methods and classical support vector classification for enhanced predictive accuracy in drug development.
• To demonstrate the feasibility of using a quantum support vector classifier with a simplified molecular input line entry system (SMILES) string kernel.

Main Methods:

• Development of a hybrid quantum-classical machine learning framework.
• Integration of a classical support vector classifier with a kernel-based quantum classifier.
• Utilized a SMILES notation-based string kernel for evaluating chemical/drug ADME-Tox properties through large-scale simulations.

Main Results:

• The proposed QML framework demonstrated superior performance compared to classical models in predicting ADME-Tox outcomes.
• Achieved high Area Under the Curve of the Receiver Operating Characteristic curve (AUC ROC) values ranging from 0.80 to 0.95.
• Performance was validated across ADME-Tox datasets for small molecules with varying feature numbers.

Conclusions:

• The developed QML framework shows significant promise for accurate and efficient prediction of crucial drug properties.
• Quantum machine learning offers a valuable tool to accelerate the drug discovery pipeline by reducing costs and development time.
• Deployment in the pharmaceutical industry could lead to more informed decision-making in R&D.