Another string to your bow: machine learning prediction of the pharmacokinetic properties of small molecules

INTRODUCTION

Prediction of pharmacokinetic (PK) properties is crucial for drug discovery and development. Machine-learning (ML) models, which use statistical pattern recognition to learn correlations between input features (such as chemical structures) and target variables (such as PK parameters), are being increasingly used for this purpose. To embed ML models for PK prediction into workflows and to guide future development, a solid understanding of their applicability, advantages, limitations, and synergies with other approaches is necessary.

AREAS COVERED

This narrative review discusses the design and application of ML models to predict PK parameters of small molecules, especially in light of established approaches including in vitro-in vivo extrapolation (IVIVE) and physiologically based pharmacokinetic (PBPK) models. The authors illustrate scenarios in which the three approaches are used and emphasize how they enhance and complement each other. In particular, they highlight achievements, the state of the art and potentials of applying machine learning for PK prediction through a comphrehensive literature review.

EXPERT OPINION

ML models, when carefully crafted, regularly updated, and appropriately used, empower users to prioritize molecules with favorable PK properties. Informed practitioners can leverage these models to improve the efficiency of drug discovery and development process.

Sorafenib reduces the production of epoxyeicosatrienoic acids and leads to cardiac injury by inhibiting CYP2J in rats

Sorafenib, an important cancer drug in clinical practice, has caused heart problems such as hypertension, myocardial infarction, and thrombosis. Although some mechanisms of sorafenib-induced cardiotoxicity have been proposed, there is still more research needed to reach a well-established definition of the causes of cardiotoxicity of sorafenib. In this report, we demonstrate that sorafenib is a potent inhibitor of the CYP2J enzyme. Sorafenib significantly inhibited the production of epoxyeicosatrienoic acids (EETs) in rat cardiac microsomes. The in vivo experimental results also showed that after the administration of sorafenib, the levels of 11,12-EET and 14,15-EET in rat plasma were significantly reduced, which was similar to the results of CYP2J gene knockout. Sorafenib decreased the levels of EETs, leading to abnormal expression of mitochondrial fusion and fission factors in heart tissue. In addition, the expression of mitochondrial energy metabolism factors (Pgc-1α, Pgc-1β, Ampk, and Sirt1) and cardiac mechanism factors (Scn5a and Prkag2) was significantly reduced, increasing the risk of arrhythmia and heart failure. Meanwhile, the increase in injury markers Anp, CK, and CK-MB further confirmed the cardiotoxicity of sorafenib. This study is of great significance for understanding the cardiotoxicity of sorafenib, and is also a model for studying the cardiotoxicity of other drugs that inhibit CYP2J activity.

Multiple Metabolism Pathways of Bentazone Potentially Regulated by Metabolic Enzymes in Rice

Bentazone (BNTZ) is a selective and efficient herbicide used in crop production worldwide. However, the persistence of BNTZ residues in the environment has led to their increasing accumulation in farmland and crops, posing a high risk to human health. To evaluate its impact on crop growth and environmental safety, a comprehensive study was conducted on BNTZ toxicity, metabolic mechanism, and resultant pathways in rice. The rice growth was compromised to the treatment with BNTZ at 0.2-0.8 mg/L (529.95-1060.05 g a.i./ha), while the activities of enzymes including SOD, POD, CAT, GST, GT, and CYP450 were elevated under BNTZ stress. A genome-wide RNA-sequencing (RNA-Seq) was performed to dissect the variation of transcriptomes and metabolic mechanisms in rice exposed to BNTZ. The degradative pathways of BNTZ in rice are involved in glycosylation, hydrolysis, acetylation, and conjugation processes catalyzed by the enzymes. Our data provided evidence that helps understand the BNTZ metabolic and detoxic mechanisms.

CYP2J deficiency leads to cardiac injury and presents dual regulatory effects on cardiac function in rats

Cytochrome P450 2 J2 (CYP2J2) enzyme is widely expressed in aortic endothelial cells and cardiac myocytes and affects cardiac function, but the underlying mechanism is still unclear. Based on CYP2J knockout (KO) rats, we have directly studied the metabolic regulation of CYP2J on cardiac function during aging. The results showed that CYP2J deficiency significantly reduced the content of epoxyeicosatrienoic acids (EETs) in plasma, aggravated myocarditis, myocardial hypertrophy, as well as fibrosis, and inhibited the mitochondrial energy metabolism signal network Pgc-1α/Ampk/Sirt1. With the increase of age, the levels of 11,12-EET and 14,15-EET in plasma of KO rats decreased significantly, and the heart injury was more serious. Interestingly, we found that after CYP2J deletion, the heart initiated a self-protection mechanism by upregulating cardiac mechanism factors Myh7, Dsp, Tnni3, Tnni2, and Scn5a, as well as mitochondrial fusion factors Mfn2 and Opa1. However, this protective effect disappeared with aging. In conclusion, CYP2J deficiency not only reduces the amount of EETs, but also plays a dual regulatory role in cardiac function.

Differentially Expressed Genes and Molecular Susceptibility to Human Age-Related Diseases

Mainstream transcriptome profiling of susceptibility versus resistance to age-related diseases (ARDs) is focused on differentially expressed genes (DEGs) specific to gender, age, and pathogeneses. This approach fits in well with predictive, preventive, personalized, participatory medicine and helps understand how, why, when, and what ARDs one can develop depending on their genetic background. Within this mainstream paradigm, we wanted to find out whether the known ARD-linked DEGs available in PubMed can reveal a molecular marker that will serve the purpose in anyone's any tissue at any time. We sequenced the periaqueductal gray (PAG) transcriptome of tame versus aggressive rats, identified rat-behavior-related DEGs, and compared them with their known homologous animal ARD-linked DEGs. This analysis yielded statistically significant correlations between behavior-related and ARD-susceptibility-related fold changes (log₂ values) in the expression of these DEG homologs. We found principal components, PC1 and PC2, corresponding to the half-sum and the half-difference of these log₂ values, respectively. With the DEGs linked to ARD susceptibility and ARD resistance in humans used as controls, we verified these principal components. This yielded only one statistically significant common molecular marker for ARDs: an excess of Fcγ receptor IIb suppressing immune cell hyperactivation.

Evaluation of Herb–Drug Interaction Between Danshen and Rivaroxaban in Rat and Human Liver Microsomes

The combination of Salvia miltiorrhiza (Danshen) and rivaroxaban is a promising treatment option in clinical practice in China, but the herb–drug interaction between Danshen and rivaroxaban remains unclear. Therefore, this study aims to reveal the interaction between Danshen and rivaroxaban. We not only investigated the inhibitory properties of Danshen tablet on rivaroxaban metabolism in rat and human liver microsomes but also evaluated the inhibitory effects of Danshen tablet and its eight active components (dihydrotanshinone I, tanshinone I, tanshinone IIA, cryptotanshinone, danshensu, salvianolic acid A, salvianolic acid B, and salvianolic acid C) on cytochrome P450 (CYP) enzymes. The results showed that Danshen tablet potently inhibited the metabolism of rivaroxaban in rat and human liver microsomes. In the CYP inhibition study, we found that dihydrotanshinone I, the active component of Danshen tablet, potently inhibited the activities of rat CYP3A and CYP2J, with IC₅₀ values at 13.85 and 6.39 μM, respectively. In further inhibition kinetic study, we found that Danshen tablet is a mixed inhibitor in rivaroxaban metabolism in rat and human liver microsomes, with the K_i value at 0.72 and 0.25 mg/ml, respectively. In conclusion, there is a potential interaction between Danshen tablet and rivaroxaban. Danshen tablet inhibits the metabolism of rivaroxaban, which may be because its lipid-soluble components such as dihydrotanshinone I strongly inhibit the activities of CYP enzymes, especially CYP3A and CYP2J. Therefore, when Danshen tablet and rivaroxaban are used simultaneously in the clinic, it is necessary to strengthen the drug monitoring of rivaroxaban and adjust the dosage.

Site-directed deuteration of dronedarone preserves cytochrome P4502J2 activity and mitigates its cardiac adverse effects in canine arrhythmic hearts

Cytochrome P4502J2 (CYP2J2) metabolizes arachidonic acid (AA) to cardioprotective epoxyeicosatrienoic acids (EETs). Dronedarone, an antiarrhythmic drug prescribed for treatment of atrial fibrillation (AF) induces cardiac adverse effects (AEs) with poorly understood mechanisms. We previously demonstrated that dronedarone inactivates CYP2J2 potently and irreversibly, disrupts AA-EET pathway leading to cardiac mitochondrial toxicity rescuable via EET enrichment. In this study, we investigated if mitigation of CYP2J2 inhibition prevents dronedarone-induced cardiac AEs. We first synthesized a deuterated analogue of dronedarone (termed poyendarone) and demonstrated that it neither inactivates CYP2J2, disrupts AA-EETs metabolism nor causes cardiac mitochondrial toxicity in vitro. Our patch-clamp experiments demonstrated that pharmacoelectrophysiology of dronedarone is unaffected by deuteration. Next, we show that dronedarone treatment or CYP2J2 knockdown in spontaneously beating cardiomyocytes indicative of depleted CYP2J2 activity exacerbates beat-to-beat (BTB) variability reflective of proarrhythmic phenotype. In contrast, poyendarone treatment yields significantly lower BTB variability compared to dronedarone in cardiomyocytes indicative of preserved CYP2J2 activity. Importantly, poyendarone and dronedarone display similar antiarrhythmic properties in the canine model of persistent AF, while poyendarone substantially reduces beat-to-beat variability of repolarization duration suggestive of diminished proarrhythmic risk. Our findings prove that deuteration of dronedarone prevents CYP2J2 inactivation and mitigates dronedarone-induced cardiac AEs.

Recent Advances in the Production of Genome-Edited Rats

The rat is an important animal model for understanding gene function and developing human disease models. Knocking out a gene function in rats was difficult until recently, when a series of genome editing (GE) technologies, including zinc-finger nucleases (ZFNs), transcription activator-like effector nucleases (TALENs), and the type II bacterial clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated Cas9 (CRISPR/Cas9) systems were successfully applied for gene modification (as exemplified by gene-specific knockout and knock-in) in the endogenous target genes of various organisms including rats. Owing to its simple application for gene modification and its ease of use, the CRISPR/Cas9 system is now commonly used worldwide. The most important aspect of this process is the selection of the method used to deliver GE components to rat embryos. In earlier stages, the microinjection (MI) of GE components into the cytoplasm and/or nuclei of a zygote was frequently employed. However, this method is associated with the use of an expensive manipulator system, the skills required to operate it, and the egg transfer (ET) of MI-treated embryos to recipient females for further development. In vitro electroporation (EP) of zygotes is next recognized as a simple and rapid method to introduce GE components to produce GE animals. Furthermore, in vitro transduction of rat embryos with adeno-associated viruses is potentially effective for obtaining GE rats. However, these two approaches also require ET. The use of gene-engineered embryonic stem cells or spermatogonial stem cells appears to be of interest to obtain GE rats; however, the procedure itself is difficult and laborious. Genome-editing via oviductal nucleic acids delivery (GONAD) (or improved GONAD (i-GONAD)) is a novel method allowing for the in situ production of GE zygotes existing within the oviductal lumen. This can be performed by the simple intraoviductal injection of GE components and subsequent in vivo EP toward the injected oviducts and does not require ET. In this review, we describe the development of various approaches for producing GE rats together with an assessment of their technical advantages and limitations, and present new GE-related technologies and current achievements using those rats in relation to human diseases.

CRISPR-Cas9: A method for establishing rat models of drug metabolism and pharmacokinetics

The 2020 Nobel Prize in Chemistry recognized CRISPR-Cas9, a super-selective and precise gene editing tool. CRISPR-Cas9 has an obvious advantage in editing multiple genes in the same cell, and presents great potential in disease treatment and animal model construction. In recent years, CRISPR-Cas9 has been used to establish a series of rat models of drug metabolism and pharmacokinetics (DMPK), such as Cyp, Abcb1, Oatp1b2 gene knockout rats. These new rat models are not only widely used in the study of drug metabolism, chemical toxicity, and carcinogenicity, but also promote the study of DMPK related mechanism, and further strengthen the relationship between drug metabolism and pharmacology/toxicology. This review systematically introduces the advantages and disadvantages of CRISPR-Cas9, summarizes the methods of establishing DMPK rat models, discusses the main challenges in this field, and proposes strategies to overcome these problems.

Characterization of a Novel CYP1A2 Knockout Rat Model Constructed by CRISPR/Cas9

CYP1A2, as one of the most important cytochrome P450 isoforms, is involved in the biotransformation of many important endogenous and exogenous substances. CYP1A2 also plays an important role in the development of many diseases because it is involved in the biotransformation of precancerous substances and poisons. Although the generation of Cyp1a2 knockout (KO) mouse model has been reported, there are still no relevant rat models for the study of CYP1A2-mediated pharmacokinetics and diseases. In this report, CYP1A2 KO rat model was established successfully by CRISPR/Cas9 without any detectable off-target effect. Compared with wild-type rats, this model showed a loss of CYP1A2 protein expression in the liver. The results of pharmacokinetics in vivo and incubation in vitro of specific substrates of CYP1A2 confirmed the lack of function of CYP1A2 in KO rats. In further studies of potential compensatory effects, we found that CYP1A1 was significantly upregulated, and CYP2E1, CYP3A2, and liver X receptor β were downregulated in KO rats. In addition, CYP1A2 KO rats exhibited a significant increase in serum cholesterol and free testosterone accompanied by mild liver damage and lipid deposition, suggesting that CYP1A2 deficiency affects lipid metabolism and liver function to a certain extent. In summary, we successfully constructed the CYP1A2 KO rat model, which provides a useful tool for studying the metabolic function and physiologic function of CYP1A2. SIGNIFICANCE STATEMENT: Human CYP1A2 not only metabolizes clinical drugs and pollutants but also mediates the biotransformation of endogenous substances and plays an important role in the development of many diseases. However, there are no relevant CYP1A2 rat models for the research of pharmacokinetics and diseases. This study successfully established CYP1A2 knockout rat model by using CRISPR/Cas9. This rat model provides a powerful tool to study the function of CYP1A2 in drug metabolism and diseases.

Recent developments in in vitro and in vivo models for improved translation of preclinical pharmacokinetics and pharmacodynamics data

Improved pharmacokinetics/pharmacodynamics (PK/PD) prediction in the early stages of drug development is essential to inform lead optimization strategies and reduce attrition rates. Recently, there have been significant advancements in the development of new in vitro and in vivo strategies to better characterize pharmacokinetic properties and efficacy of drug leads. Herein, we review advances in experimental and mathematical models for clearance predictions, advancements in developing novel tools to capture slowly metabolized drugs, in vivo model developments to capture human etiology for supporting drug development, limitations and gaps in these efforts, and a perspective on the future in the field.