New Machine Learning Models for Predicting the Organic Cation Transporters OCT1, OCT2, and OCT3 Uptake

Area of Science:

• Pharmacology
• Computational Chemistry
• Biochemistry

Background:

• Organic cation transporters (OCTs) are crucial for drug pharmacokinetics (PK), influencing uptake and elimination.
• Current experimental methods for assessing OCT modulation by small molecules are costly and time-consuming.
• Accurate in silico models are needed for early prediction of PK liabilities in drug candidates.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting the uptake of organic cation transporters 1, 2, and 3 (OCT1, OCT2, OCT3).
• To provide a cost-effective computational tool for forecasting drug interactions with OCTs.
• To aid in the early identification of potential PK liabilities during drug discovery.

Main Methods:

• Utilized advanced decision tree ensemble algorithms for model construction.
• Employed VolSurf molecular descriptors as features for the machine learning models.
• Trained and validated models on the largest available curated datasets for OCT-mediated uptake.

Main Results:

• Developed ML models with significant predictive power for OCT1, OCT2, and OCT3 uptake.
• Achieved Matthews correlation coefficient (MCC) values exceeding 0.45 in multiple external validation rounds.
• Demonstrated the robustness and accuracy of the developed in silico models.

Conclusions:

• The developed ML models offer a cost-effective alternative to experimental assays for predicting OCT-mediated drug uptake.
• These models can significantly aid drug discovery by enabling early forecasting of PK liabilities.
• The findings provide valuable tools for understanding and mitigating OCT-related challenges in drug development.