EUROTOX 2024 – 58th Congress of European Societies of Toxicology

K. Lanevskij (1,2), R. Didžiapetris (1,2), A. Sazonovas (1,2)

(1) ACD/Labs, Inc., Toronto, Ontario, Canada
(2) VšĮ Aukštieji Algoritmai, Vilnius, Lithuania

Human P-glycoprotein (P-gp) is a major carrier protein expressed in a variety of tissues including the blood-brain barrier. P-gp is responsible for removal of xenobiotics from the cells and protecting the tissues from the potentially toxic action  of chemicals. At the same time P-gp efflux may preclude delivery of pharmaceuticals to their site of action and contribute to loss of efficacy. Numerically, P-gp effect is commonly described by the compound’s efflux ratio (ER), i.e., the ratio of BA to AB permeation rates in polarized transport assay s. Due to the lack of accurate quantitative measurements, computational studies traditionally treat P-gp efflux as a binary endpoint and only attempt to classify molecules as P-gp substrates or non-substrates.
However, recently we have proposed a QSAR model that can produce quantitative predictions of ER values [1]. The model was parameterized using a minimal set of key physicochemical descriptors (logD, pKa, molecular size, etc.) and a censored regression-based machine learning methodology that can use experimental data characterized as either exact data points or ER intervals (censored values). In the current study we extend this approach by applying an estimate of passive permeability in Caco-2 cells [2] to split measured ER values into the contributions of passive and active transport routes, and subsequently fitting the model to represent pure P-gp efflux effect. The new approach enabled us achieving similar predictivity on the qualitative classification task (> 75% overall accuracy at a threshold of ER > 2 for substrates), while having better interpretability compared to the previous model. Practical utility of quantitative predictions is demonstrated by incorporating predicted ER values into the model characterizing drugs’ accessibility to CNS [3]. The respective model supplemented with P-gp efflux estimates was able to predict one of the key brain penetration characteristics, the unbound brain/blood distribution ratio (Kp,uu) with R2 > 0.5.

[1] Lanevskij Kiril, Didziapetris Remigijus, Sazonovas Andrius 2023, ‘Physicochemical QSAR Model of P-glycoprotein Efflux Ratio Based on Quantitative and Censored Data’, Toxicol Lett, 384, S118.
[2] Lanevskij Kiril, Didziapetris Remigijus 2019, ‘Physicochemical QSAR Analysis of Passive Permeability Across Caco-2 Monolayers’, J Pharm Sci, 108, 78.
[3] Lanevskij K, Japertas P, Didziapetris R, Petrauskas A 2009, ‘Ionization-specific QSAR models of blood-brain penetration of drugs’, Chem Biodivers, 6, 2050.