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

Organic cation transporters
(OCTs) are a small family of transmembrane
proteins that regulate the pharmacokinetics (PK) of natural metabolites
and xenobiotics by facilitating drug uptake and elimination. Measuring
the modulation (either inhibition or substrate) of OCTs by small molecules
requires expensive experiments. More cost-effective in silico models that accurately predict OCT-mediated uptake would enable
the forecasting of potential PK liabilities of new drug candidates
at an early stage. In this paper, we present new machine learning
(ML) models to predict the uptake of OCT1, OCT2, and OCT3. Built using
advanced decision tree ensemble algorithms and VolSurf molecular features,
these models are based on the largest and most well-curated data sets
available in the current literature. Several rounds of validation
with different external test sets have confirmed the predictive power
of these models, with Matthews correlation coefficient (MCC) values
above 0.45. We believe that these models will shed new light on the
impact of OCTs on drug discovery and development.