Reported liver toxicity of food chemicals in rats extrapolated to humans using virtual human-to-rat hepatic concentration ratios generated by pharmacokinetic modeling with machine learning-derived parameters

Abstract

Pharmacokinetic data are not generally available for evaluating the toxicological potential of food chemicals. A simplified physiologically based pharmacokinetic (PBPK) model has been established to evaluate internal exposures to chemicals in rats or humans with no reference to in vitro or in vivo experimental data. In this study, reported liver toxicity levels in rats were extrapolated to humans using virtual hepatic concentration–time curves (AUC) as the interspecies factor. Virtual liver exposures to 27 lipophilic food chemicals (octanol–water partition coefficient logP >1) with reported rat hepatic lowest-observed-effect levels (LOELs) of ≤1000 mg/kg/day were generated using PBPK models with input parameters obtained entirely in silico via machine learning algorithms. The resulting virtual rat and human liver AUCs were correlated (n = 27, r = 0.52, p < 0.01). However, AUCs for the phenolic compounds emodin, isoeugenol, and tert-butylhydroquinone, which have reported rat LOEL values of ≤300 mg/kg/day, were located outside the relatively wide 95% confidence interval, indicating more extensive hepatic elimination in rats than in humans. In vitro depletion of tert-butylhydroquinone in rat liver fractions via sulfation was confirmed to be faster than that in humans. For emodin, isoeugenol, and tert-butylhydroquinone, human-to-rat AUC ratios ranged from 10- to 13-fold; consequently, their extrapolated human hepatic LOEL values were estimated as ≤30 mg/kg/day, i.e., one order of magnitude smaller than the rat LOELs. Despite the small number of lipophilic food chemicals considered here, the PBPK modeling approach using in silico-generated input parameters for rats and humans has the potential to facilitate toxicological studies.

INTRODUCTION

High-throughput in vitro assays in combination with in silico computational models could provide alternatives to animal testing, thereby reducing the use of experimental animals when estimating the health risks of a wide range of chemical substances (Wambaugh et al., 2018; Krewski et al., 2020). Under the Japanese Chemical Substances Control Law, individual chemicals are classified into hazard classes 1–4, based on key oral doses after 28-day repeated-dose toxicity studies in rats (Igarashi et al., 2018a, 2018b), e.g., a cut off of 30 mg/kg body weight/day as the no observed adverse effect level or the lowest observed adverse effect level. When evaluating the toxicological potential of industrial or food chemicals, pharmacokinetic and/or toxicokinetic data are generally not considered (Bell et al., 2018; Ellison et al., 2019). However, a simplified physiologically based pharmacokinetic (PBPK) model has been established that uses in silico input parameters to evaluate the internal exposures of chemicals in rats or humans (Kamiya et al., 2019; Tan et al., 2020); currently, these models can be applied with no reference to in vitro or in vivo experimental data (Kamiya et al., 2021b, 2021a, 2022a). This has been achieved by using machine learning algorithms on up to 29 in silico chemical descriptors to obtain purely in silico estimations of the input parameters for rat and human PBPK models. The four main input parameters for these models are the fraction absorbed × intestinal availability, F_aF_g; the absorption rate constant, k_a; the volume of the systemic circulation, V₁; and the hepatic intrinsic clearance, CL_h,int; recently, the machine learning algorithms were trained on expanded panels of 372 (Adachi et al., 2022a) and 355 (Adachi et al., 2022b) chemicals for rat and human PBPK models, respectively.

To assess the safety of agricultural chemicals, improved studies should be scientifically appropriate and necessary, without being redundant for toxicological endpoints and exposure durations that are relevant for risk assessment (Barton et al., 2006). We previously demonstrated that the areas under the hepatic concentration–time curves (AUC) of seven compounds estimated using rat PBPK modeling were inversely associated with the reported lowest-observed-effect levels (LOEL) available for hepatotoxicity on the Hazard Evaluation Support System Integrated Platform in Japan (Kamiya et al., 2020). To expand the scope of this approach, the hepatic LOEL levels (mg/kg body weight/day) in rats of 56 food additives and functional food ingredients after repeated oral administrations (mostly lasting 4 weeks, but with some lasting 13 weeks) were obtained from previous reports (Adachi et al., 2023a). Using rat PBPK forward dosimetry with in silico input parameters provided the plasma/hepatic concentrations of these 56 substances. However, the correlation between hepatic AUCs after oral dosing and LOEL levels was significant only for 14 hydrophobic food chemicals (octanol–water partition coefficient logP >1). Nonetheless, this approach could provide estimates of hepatotoxic potential as part of the risk assessment of these substances (Adachi et al., 2023a).

When calculating logP values using software such as ChemDraw or ACD/Lab Percepta, some differences were observed depending on the software package used (Mannhold et al., 2009). We therefore suggest that an impact of variations in in silico/in vitro logP values on various pharmacokinetic parameters be carried out (Adachi et al., 2024b). The aim of the present study was to extrapolate reported liver toxicity data in rats to humans using PBPK-generated virtual hepatic AUC values of food chemicals as the interspecies factor (Fig. 1). We report herein that there was a tendency for the PBPK-modeled internal exposures to lipophilic food chemicals in rats and humans to be inversely associated with the rat LOEL values. However, some phenolic compounds, such as isoeugenol, emodin, and tert-butylhydroquinone, which have reported rat LOEL values of <300 mg/kg/day, exhibited more extensive hepatic elimination in rats than in humans; consequently, under the present conditions, their LOELs as extrapolated to humans were one-order of magnitude smaller at ≤30 mg/kg/day, which, based on their toxic potential, implies a one-grade change of hazard class classification. Based on estimating interspecies toxicokinetics or internal exposures of lipophilic food components after oral doses in humans, this approach, which applies simple PBPK modeling with no reference to experimental pharmacokinetic data, has the potential to play a significant role in the extrapolation of reported liver toxicity levels in rats to humans.

Fig. 1

Schematic of method for extrapolating reported liver toxicity data in rats to humans using virtual liver AUC ratios after oral 1.0 mg/kg doses as the interspecies factor. PBPK models were used with input pharmacokinetic parameters derived completely in silico by a machine-learning algorithm.

MATERIALS AND METHODS

In silico analysis

To estimate the input parameters for PBPK modeling using a previously established machine learning algorithm that solely utilizes in silico chemical descriptors (Adachi et al., 2022a, 2022b), the logP values and acid dissociation constants of 27 hydrophobic food chemicals with rat LOEL values <1000 mg/kg/day were obtained using ACD/Percepta software (Advanced Chemistry Development, Toronto, ON, Canada) (Adachi et al., 2024b). To visualize and confirm a variety of their chemical structures, the chemical space represented by the food chemicals was projected onto a two-dimensional plane using generative topographic mapping (Adachi et al., 2023a). The fraction of the unionized form (designated f_mol in this study) of each compound at the experimental pH conditions was estimated with the Henderson–Hasselbalch equation using the respective pK_a value (Kamiya et al., 2021c). The plasma unbound fractions (f_u,p) of the test compounds were independently estimated for rats and humans using Simcyp software (Certara, Sheffield, UK). The in silico PBPK model input parameters (F_aF_g, k_a, V₁, and CL_h,int) were estimated using an updated machine learning algorithm trained with 372 (Adachi et al., 2022a) and 355 (Adachi et al., 2022b) diverse chemicals for rats and humans, respectively. The liver-to-plasma partition coefficients were calculated using the Poulin and Theil equations (Adachi et al., 2023b).

To generate the plasma/hepatic AUCs from 0 to 24 hr after single virtual oral 1.0 mg/kg doses of the 27 selected substances, sets of differential equations representing simplified rat and human PBPK models were solved as reported previously (Kamiya et al., 2021b; Adachi et al., 2024a). Correlations between hepatic LOEL levels (mg/kg/day) in rats and liver AUC values (ng hr/g liver) in rats and humans were evaluated using Prism 10 (GraphPad Software, La Jolla, CA, USA) for linear regression and 95% confidence intervals.

In vitro analysis

To empirically investigate one of the in silico PBPK forward dosimetry findings, in vitro rat and human hepatic sulfation-dependent substrate depletion assays were performed. Tert-butylhydroquinone (10 µM, Tokyo Chemical Industry, Tokyo, Japan) was incubated with liver 9000 × g supernatant fractions from humans and rats (S9, 0.50 mg/mL, XenoTech, Kansas City, KS, USA) in 100 mM potassium phosphate buffer (pH 7.4) containing 1.0 mM adenosine 3'-phosphoadenosine-5'-phosphosulfate (PAPS, Sigma-Aldrich, St. Louis, MO, USA) and an NADPH-regenerating system in a total volume of 0.25 mL. All samples were incubated at 37°C for 0–30 min, quenched by adding 0.25 mL of methanol, and centrifuged at 900 × g. Supernatant (30 µL) was injected into a high-performance liquid chromatography system equipped with an analytical octadecylsilane column. The remaining tert-butylhydroquinone was analyzed using an isocratic eluent [70% (v/v) methanol in 20 mM phosphoric acid] at a flow rate of 1.0 mL/min at 40°C under UV detection at 280 nm.

RESULTS

The PBPK model input parameters for the 27 hydrophobic food chemicals were calculated using a previously established machine learning algorithm and are shown in Table 1, along with physicochemical values, including logP, obtained using ACD/Percepta software. V₁ and CL_h,int values for rat and human PBPK modeling are shown after being normalized using body weights for rats (0.25 kg) and humans (70 kg). Figure 2A shows the tendency for inverse correlation between reported logarithm-transformed hepatic LOEL values in rats and modeled human liver AUC values after virtual oral doses of 1.0 mg/kg. There was no significant relationship between rat LOELs (indicating hepatic toxicity) and virtual human liver AUCs for the 27 compounds with LOEL values <1000 mg/kg/day (r = –0.27, p = 0.17) or for the 17 substances with reported LOEL values <300 mg/kg/day (r = –0.31, p = 0.22). However, there was a significant correlation (n = 27, r = 0.52, p < 0.01) between the logarithm-transformed PBPK model-generated hepatic AUC values for rats and humans after virtual oral doses of 1.0 mg/kg (Fig. 2B). For the 17 chemicals with LOEL values of <300 mg/kg/day, indicating more potent hepatotoxicity, a significant linear regression line (r = 0.70, p < 0.01) with a relatively wide 95% confidence interval is illustrated (Fig. 2B, gray solid and dotted lines). For five compounds (emodin, isoeugenol, tert-butylhydroquinone, senkirkine, and harman), the hepatic AUC values in human liver were considerably higher than those the rat liver (Fig. 2B). The former three compounds are common phenolic derivatives. In the narrower 95% confidence intervals for LOEL values <1000 mg/kg/day, three additional compounds (curcumin, lasiocarpine, and citral) had higher hepatic AUC values in human liver than in rat liver. Curcumin also contains phenolic hydroxyl groups.

Table 1. Input parameters for rat and human PBPK models for 27 hydrophobic food chemicals predicted by machine learning systems.

Fig. 2

Relationship between reported hepatic LOEL values in rats and virtual AUC values in human liver after oral doses of 1.0 mg/kg (A) and between virtual AUC values in rat and human livers (B) for 27 hydrophobic food chemicals. The AUC values in human liver were generated by human PBPK models using the pharmacokinetic parameters shown in Table 1. Panel B shows the regression and 95% confidence interval lines (solid and dotted lines, respectively) for 17 chemicals with hepatic LOEL values ≤300 mg/kg/day (gray circles and gray lines) and 27 chemicals with hepatic LOEL values ≤1000 mg/kg/day (open circles and black lines).

The virtual plasma and liver concentration profiles in rats and humans after a virtual oral dose of 1.0 mg/kg of isoeugenol, emodin, and tert-butylhydroquinone, along with their chemical structures, are shown in Fig. 3. These compounds, particularly tert-butylhydroquinone, showed more rapid clearance from the plasma and liver in rats than in humans (Fig. 3C). The AUC values for rat and human livers are summarized in Table 2. For these three substances, the virtual AUC values in human liver were higher than those in rat liver; the ratios of the human to rat hepatic AUC values ranged from 10- to 13-fold (Table 2).

Fig. 3

Liver and plasma concentrations of isoeugenol (A), emodin (B), and tert-butylhydroquinone (C) in rats and humans estimated using PBPK models. Liver (dashed lines) and plasma (solid lines) concentrations were derived after a virtual oral dose of 1.0 mg/kg and are shown for rats (blue) and humans (green).

Table 2. Extrapolated human hepatic LOEL values of selected food chemicals using virtual hepatic exposure ratios generated by human and rat PBPK models as the conversion factor.

In vitro rat and human hepatic sulfation-dependent substrate depletion assays were performed for tert-butylhydroquinone (Fig. 4). For 30 min incubations, the depletion rates of tert-butylhydroquinone (at 10 μM) in liver S9 fractions from rats and humans were calculated as 0.59 and 0.13 nmol/min/mg protein, respectively, in the presence of an NADPH-regenerating system and PAPS.

Fig. 4

Metabolic clearance profiles of tert-butylhydroquinone in liver S9 substrate-depletion assays in rats and humans. Tert-Butylhydroquinone (10 µM) was incubated with rat (blue) and human (green) liver S9 fractions in the presence of an NADPH-generating system and PAPS. Plotted data are the means and standard deviations of triplicate determinations.

DISCUSSION

Species differences may cause problems when using in vitro assays in combination with in silico computational models for animal testing of the health risks of chemical substances (Wambaugh et al., 2018; Krewski et al., 2020). The Organization for Economic Cooperation and Development (OECD) published a guidance document describing a scientific workflow for characterizing and validating PBPK models developed using in vitro and in silico data (Paini et al., 2021). In that document with case studies, human PBPK modeling generally required actual pharmacokinetic data of chemicals reported or performed in rats or humanized-liver mice (Takano et al., 2021; Miura et al., 2021). Recently, these models have been applied with no reference to in vitro or in vivo experimental data (Kamiya et al., 2021b, 2021a, 2022a). In the current study, a significant but relatively small correlation coefficient (n = 27, r = 0.52, p < 0.01) was observed between the logarithm-transformed hepatic AUC values in rat and human livers modeled after virtual oral doses of 1.0 mg/kg (Fig. 2B). In terms of the liver-to-plasma partition coefficients in this study, we previously confirmed that values derived from the primary Poulin and Theil model could be used to estimate internal exposure (Adachi et al., 2023b) in comparison with those of the Rodgers and Rowland method in recent surveys of 14 medicinal substances in rats, even though there are limited experimental plasma and liver concentration data available for which future work would be expected. These in silico computational models could provide international alternatives to reduce animal experimental testing of a wide range of chemical substances.

Many factors may contribute to these species differences. For example, rats are reported generally to exhibit higher drug metabolizing activities than humans, including for sulfation (Gamage et al., 2006). In this study, we focused on the 17 chemicals with more potentially toxic LOEL values of <300 mg/kg/day; the plots for the phenolic derivatives emodin, isoeugenol, and tert-butylhydroquinone fell in the region where human hepatic AUC values were considerably higher than those for rat liver (Fig. 2B). More rapid in silico-estimated clearances of isoeugenol, emodin, and tert-butylhydroquinone from rat plasma and liver, compared with those in humans, were evident (Fig. 3). Tert-butylhydroquinone is reportedly eliminated by sulfation during in vitro incubation with human and rat liver S9 fractions (Ousji and Sleno, 2020). In the present study, in the presence of an NADPH-regenerating system and PAPS, in vitro depletion of tert-butylhydroquinone in liver S9 fractions from rats was faster than that for humans (Fig. 4).

Figure 1 shows a schematic of the current method of extrapolating reported liver toxicity data from rats to humans using solely in silico PBPK-modeled human-to-rat liver AUC ratios after virtual oral 1.0 mg/kg administrations as the interspecies factor. The virtual AUC values of emodin, isoeugenol, and tert-butylhydroquinone in human liver were higher than those in rat liver, with human:rat AUC ratios in the range 10- to 13-fold. Using these hepatic AUC ratios as interspecies conversion factors, the extrapolated human hepatic LOEL values of emodin, isoeugenol, and tert-butylhydroquinone were estimated to be as low as 2.9–17 mg/kg/day (Table 2). This approach potentially could be extended to other phenolic compounds with similar metabolic profiles, which may also be eliminated faster in rats than in humans.

The machine learning algorithms to generate entirely in silico input parameters for rat and human PBPK models were trained separately using data from 372 (Adachi et al., 2022a) and 355 (Adachi et al., 2022b) diverse chemicals. PBPK modeling with these in silico parameters could predict species-dependent metabolic clearances from the plasma and liver of some of the hydrophobic phenolic food compounds investigated in this study (Fig. 3). It has been noted in fish toxicity screening systems that chemicals with logP ≤ 0 (i.e., hydrophylic substances) tend to be nontoxic (Padilla et al., 2012), and that a full logD-based model employing logP as a membrane passage descriptor (Köhler et al., 2023) can be used to predict potential toxicities. However, in a preliminary study, we found similar rates of in vitro substrate depletion for harman (with a LOEL value <300 mg/kg/day) with rat and human liver S9 fractions in the presence of an NADPH-generating system and acetyl CoA; nonetheless, in the current study, harman exhibited much higher virtual hepatic AUC values for human liver than for rat liver. The in vitro senkirkine elimination rates in liver microsomes from rats and humans fortified with an NADPH-generating system were roughly similar (Kamiya et al., 2022b). Many factors may contribute to these species differences, but there may currently be unknown reasons for harman or senkirkine elimination. Recently, programmatic access and pathway management for xenobiotic metabolism simulators has been reportedly available (Groff et al., 2024). Although it should be noted that the number of lipophilic food chemicals with reported moderate LOEL values (≤300 mg/kg/day) investigated in the current study is limited, the simple PBPK modeling approach used here, i.e., with input parameters generated with no reference to experimental pharmacokinetic data, has the potential to play a significant role in reducing the use of animals for estimating the toxicokinetics or internal exposure of lipophilic food components, as well as other general industrial chemicals, after oral doses.

In conclusion, the present results suggest that some compounds with LOEL values ≤300 mg/kg/day in rats show more extensive hepatic elimination in rats than in humans. In humans, one-order smaller LOEL values of ≤30 mg/kg/day, which would imply a one-grade change of hazard class classification based on toxic potential, were derived in this extrapolation study; this finding should be considered in the future risk assessment of some phenolic compounds. The approach applied here, reinforcing some limitations or uncertainties that use simple PBPK modeling with no reference to experimental pharmacokinetic data, has the potential to play a significant role in the extrapolation of reported liver toxicity levels in rats to humans using estimated interspecies toxicokinetics/internal exposures of lipophilic food components after virtual oral doses.

ACKNOWLEDGMENTS

This study was supported in part by the Japan Society for the Promotion of Science Grant-in-Aid for Scientific Research (23K14393 and 23K06217), the METI Artificial Intelligence-based Substance Hazard Integrated Prediction System Project, Japan, and the Japan Chemical Industry Association Long-Range Research Initiative Program. We thank Certara UK (Simcyp Division) for providing academic access to simulators and Drs. Kimoto Funatsu, Fumiaki Shono, Masato Kitajima, Tsubasa Sasaki, Atsuo Arai and Norie Murayama for their assistance. We are also grateful to David Smallbones for copyediting a draft of this article.

Conflict of interest

The authors declare that there is no conflict of interest.

REFERENCES

•  Adachi,  K.,  Nakano,  H.,  Sato,  T.,  Shimizu,  M. and  Yamazaki,  H. (2023a): Liver and Plasma Concentrations of Food Chemicals after Virtual Oral Doses Extrapolated Using in Silico Estimated Input Pharmacokinetic Parameters to Confirm Reported Liver Toxicity in Rats. Biol. Pharm. Bull., 46, 1133-1140.
•  Adachi,  K.,  Sasaki,  T.,  Arai,  A.,  Shimizu,  M. and  Yamazaki,  H. (2024a): Impact of variability of in silico and in vitro octanol/water partition coefficients of compounds on the input parameters and results of simplified human physiologically based pharmacokinetic models after virtual oral administrations. J. Toxicol. Sci., 49, 459-466.
•  Adachi,  K.,  Shimizu,  M.,  Shono,  F.,  Funatsu,  K. and  Yamazaki,  H. (2024b): Octanol/water partition coefficients estimated using retention times in reverse-phase liquid chromatography and calculated in silico as one of the determinant factors for pharmacokinetic parameter estimations of general chemical substances. J. Toxicol. Sci., 49, 127-137.
•  Adachi,  K.,  Shimizu,  M. and  Yamazaki,  H. (2022a): Updated in silico prediction methods for fractions absorbed and absorption rate constants of 372 disparate chemicals for use in physiologically based pharmacokinetic models for estimating internal concentrations in rats. J. Toxicol. Sci., 47, 453-456.
•  Adachi,  K.,  Shimizu,  M. and  Yamazaki,  H. (2022b): Updated in Silico Prediction Methods for Fractions Absorbed and Key Input Parameters of 355 Disparate Chemicals for Physiologically Based Pharmacokinetic Models for Time-Dependent Plasma Concentrations after Virtual Oral Doses in Humans. Biol. Pharm. Bull., 45, 1812-1817.
•  Adachi,  K.,  Utsumi,  M.,  Sato,  T.,  Nakano,  H.,  Shimizu,  M. and  Yamazaki,  H. (2023b): Modeled Rat Hepatic and Plasma Concentrations of Chemicals after Virtual Administrations Using Two Sets of in Silico Liver-to-Plasma Partition Coefficients. Biol. Pharm. Bull., 46, 1316-1323.
•  Barton,  H.A.,  Pastoor,  T.P.,  Baetcke,  K.,  Chambers,  J.E.,  Diliberto,  J.,  Doerrer,  N.G.,  Driver,  J.H.,  Hastings,  C.E.,  Iyengar,  S.,  Krieger,  R.,  Stahl,  B. and  Timchalk,  C. (2006): The acquisition and application of absorption, distribution, metabolism, and excretion (ADME) data in agricultural chemical safety assessments. Crit. Rev. Toxicol., 36, 9-35.
•  Bell,  S.M.,  Chang,  X.,  Wambaugh,  J.F.,  Allen,  D.G.,  Bartels,  M.,  Brouwer,  K.L.,  Casey,  W.M.,  Choksi,  N.,  Ferguson,  S.S.,  Fraczkiewicz,  G.,  Jarabek,  A.M.,  Ke,  A.,  Lumen,  A.,  Lynn,  S.G.,  Paini,  A.,  Price,  P.S.,  Ring,  C.,  Simon,  T.W.,  Sipes,  N.S.,  Sprankle,  C.S.,  Strickland,  J.,  Troutman,  J.,  Wetmore,  B.A. and  Kleinstreuer,  N.C. (2018): In vitro to in vivo extrapolation for high throughput prioritization and decision making. Toxicol. In Vitro, 47, 213-227.
•  Ellison,  C.A.,  Blackburn,  K.L.,  Carmichael,  P.L.,  Clewell,  H.J. 3rd,  Cronin,  M.T.,  Desprez,  B.,  Escher,  S.E.,  Ferguson,  S.S.,  Grégoire,  S.,  Hewitt,  N.J.,  Hollnagel,  H.M.,  Klaric,  M.,  Patel,  A.,  Salhi,  S.,  Schepky,  A.,  Schmitt,  B.G.,  Wambaugh,  J.F. and  Worth,  A. (2019): Challenges in working towards an internal threshold of toxicological concern (iTTC) for use in the safety assessment of cosmetics: Discussions from the Cosmetics Europe iTTC Working Group workshop. Regul. Toxicol. Pharmacol., 103, 63-72.
•  Gamage,  N.,  Barnett,  A.,  Hempel,  N.,  Duggleby,  R.G.,  Windmill,  K.F.,  Martin,  J.L. and  McManus,  M.E. (2006): Human sulfotransferases and their role in chemical metabolism. Toxicol. Sci., 90, 5-22.
•  Groff,  L.,  Williams,  A.,  Shah,  I. and  Patlewicz,  G. (2024): MetSim: Integrated Programmatic Access and Pathway Management for Xenobiotic Metabolism Simulators. Chem. Res. Toxicol., 37, 685-697.
•  Igarashi,  T.,  Serizawa,  H.,  Kobayashi,  K.,  Suzuki,  H.,  Matsumoto,  M.,  Iso,  T.,  Kawamura,  T.,  Inoue,  K.,  Ono,  A.,  Yamada,  T. and  Hirose,  A. (2018a): Initial hazard assessment of 4-benzylphenol, a structural analog of bisphenol F: genotoxicity tests in vitro and a 28-day repeated-dose toxicity study in rats. Regul. Toxicol. Pharmacol., 96, 64-75.
•  Igarashi,  T.,  Takashima,  H.,  Takabe,  M.,  Suzuki,  H.,  Ushida,  K.,  Kawamura,  T.,  Matsumoto,  M.,  Iso,  T.,  Tanabe,  S.,  Inoue,  K.,  Ono,  A.,  Yamada,  T. and  Hirose,  A. (2018b): Initial hazard assessment of benzyl salicylate: in vitro genotoxicity test and combined repeated-dose and reproductive/developmental toxicity screening test in rats. Regul. Toxicol. Pharmacol., 100, 105-117.
•  Kamiya,  Y.,  Handa,  K.,  Miura,  T.,  Ohori,  J.,  Kato,  A.,  Shimizu,  M.,  Kitajima,  M. and  Yamazaki,  H. (2022a): Machine Learning Prediction of the Three Main Input Parameters of a Simplified Physiologically Based Pharmacokinetic Model Subsequently Used to Generate Time-Dependent Plasma Concentration Data in Humans after Oral Doses of 212 Disparate Chemicals. Biol. Pharm. Bull., 45, 124-128.
•  Kamiya,  Y.,  Handa,  K.,  Miura,  T.,  Ohori,  J.,  Shimizu,  M.,  Kitajima,  M.,  Shono,  F.,  Funatsu,  K. and  Yamazaki,  H. (2021a): An Updated In Silico Prediction Method for Volumes of Systemic Circulation of 323 Disparate Chemicals for Use in Physiologically Based Pharmacokinetic Models to Estimate Plasma and Tissue Concentrations after Oral Doses in Rats. Chem. Res. Toxicol., 34, 2180-2183.
•  Kamiya,  Y.,  Handa,  K.,  Miura,  T.,  Yanagi,  M.,  Shigeta,  K.,  Hina,  S.,  Shimizu,  M.,  Kitajima,  M.,  Shono,  F.,  Funatsu,  K. and  Yamazaki,  H. (2021b): In Silico Prediction of Input Parameters for Simplified Physiologically Based Pharmacokinetic Models for Estimating Plasma, Liver, and Kidney Exposures in Rats after Oral Doses of 246 Disparate Chemicals. Chem. Res. Toxicol., 34, 507-513.
•  Kamiya,  Y.,  Miura,  T.,  Kato,  A.,  Murayama,  N.,  Shimizu,  M. and  Yamazaki,  H. (2022b): Plasma Concentration Profiles for Hepatotoxic Pyrrolizidine Alkaloid Senkirkine in Humans Extrapolated from Rat Data Sets Using a Simplified Physiologically Based Pharmacokinetic Model. Drug Metab. Bioanal. Lett., 15, 64-69.
•  Kamiya,  Y.,  Omura,  A.,  Hayasaka,  R.,  Saito,  R.,  Sano,  I.,  Handa,  K.,  Ohori,  J.,  Kitajima,  M.,  Shono,  F.,  Funatsu,  K. and  Yamazaki,  H. (2021c): Prediction of permeability across intestinal cell monolayers for 219 disparate chemicals using in vitro experimental coefficients in a pH gradient system and in silico analyses by trivariate linear regressions and machine learning. Biochem. Pharmacol., 192, 114749.
•  Kamiya,  Y.,  Otsuka,  S.,  Miura,  T.,  Takaku,  H.,  Yamada,  R.,  Nakazato,  M.,  Nakamura,  H.,  Mizuno,  S.,  Shono,  F.,  Funatsu,  K. and  Yamazaki,  H. (2019): Plasma and Hepatic Concentrations of Chemicals after Virtual Oral Administrations Extrapolated Using Rat Plasma Data and Simple Physiologically Based Pharmacokinetic Models. Chem. Res. Toxicol., 32, 211-218.
•  Kamiya,  Y.,  Otsuka,  S.,  Miura,  T.,  Yoshizawa,  M.,  Nakano,  A.,  Iwasaki,  M.,  Kobayashi,  Y.,  Shimizu,  M.,  Kitajima,  M.,  Shono,  F.,  Funatsu,  K. and  Yamazaki,  H. (2020): Physiologically Based Pharmacokinetic Models Predicting Renal and Hepatic Concentrations of Industrial Chemicals after Virtual Oral Doses in Rats. Chem. Res. Toxicol., 33, 1736-1751.
•  Köhler,  H.R.,  Gräff,  T.,  Schweizer,  M.,  Blumhardt,  J.,  Burkhardt,  J.,  Ehmann,  L.,  Hebel,  J.,  Heid,  C.,  Kundy,  L.,  Kuttler,  J.,  Malusova,  M.,  Moroff,  F.M.,  Schlösinger,  A.F.,  Schulze-Berge,  P.,  Panagopoulou,  E.I.,  Damalas,  D.E.,  Thomaidis,  N.S.,  Triebskorn,  R.,  Maletzki,  D.,  Kühnen,  U. and  von der Ohe,  P.C. (2023): LogD-based modelling and ΔlogD as a proxy for pH-dependent action of ionizable chemicals reveal the relevance of both neutral and ionic species for fish embryotoxicity and possess great potential for practical application in the regulation of chemicals. Water Res., 235, 119864.
•  Krewski,  D.,  Andersen,  M.E.,  Tyshenko,  M.G.,  Krishnan,  K.,  Hartung,  T.,  Boekelheide,  K.,  Wambaugh,  J.F.,  Jones,  D.,  Whelan,  M.,  Thomas,  R.,  Yauk,  C.,  Barton-Maclaren,  T. and  Cote,  I. (2020): Toxicity testing in the 21st century: progress in the past decade and future perspectives. Arch. Toxicol., 94, 1-58.
•  Mannhold,  R.,  Poda,  G.I.,  Ostermann,  C. and  Tetko,  I.V. (2009): Calculation of molecular lipophilicity: state-of-the-art and comparison of log P methods on more than 96,000 compounds. J. Pharm. Sci., 98, 861-893.
• Miura, T., Suemizu, H., Goto, M., Sakai, N., Iwata, H., Shimizu, M. and Yamazaki, H. (2021): PBK model predictions for monoisononyl phthalate. In: Annex 4. Cases studies to Guidance Document on the Characterisation, Validation and Reporting of PBK Models for Regulatory Purpose, OECD Environment, Health and Safety Publications Series on Testing and Assessment no. 331), pp. 95-105.
•  Ousji,  O. and  Sleno,  L. (2020): Identification of In Vitro Metabolites of Synthetic Phenolic Antioxidants BHT, BHA, and TBHQ by LC-HRMS/MS. Int. J. Mol. Sci., 21, 9525.
•  Padilla,  S.,  Corum,  D.,  Padnos,  B.,  Hunter,  D.L.,  Beam,  A.,  Houck,  K.A.,  Sipes,  N.,  Kleinstreuer,  N.,  Knudsen,  T.,  Dix,  D.J. and  Reif,  D.M. (2012): Zebrafish developmental screening of the ToxCast™ Phase I chemical library. Reprod. Toxicol., 33, 174-187.
•  Paini,  A.,  Tan,  Y.M.,  Sachana,  M. and  Worth,  A. (2021): Gaining acceptance in next generation PBK modelling approaches for regulatory assessments - An OECD international effort. Comput. Toxicol., 18, 100163.
• Takano, R., Murayama, N., Horiuchi, K., Kitajima, M., Kumamoto, M., Shono, F. and Yamazaki, H. (2021): Physiologically based pharmacokinetic (PBK) model for acrylonitrile in humans. In: Annex 4. Cases studies to Guidance Document on the Characterisation, Validation and Reporting of PBK Models for Regulatory Purpose, OECD Environment, Health and Safety Publications Series on Testing and Assessment no. 331), pp. 85-94.
•  Tan,  Y.M.,  Chan,  M.,  Chukwudebe,  A.,  Domoradzki,  J.,  Fisher,  J.,  Hack,  C.E.,  Hinderliter,  P.,  Hirasawa,  K.,  Leonard,  J.,  Lumen,  A.,  Paini,  A.,  Qian,  H.,  Ruiz,  P.,  Wambaugh,  J.,  Zhang,  F. and  Embry,  M. (2020): PBPK model reporting template for chemical risk assessment applications. Regul. Toxicol. Pharmacol., 115, 104691.
•  Wambaugh,  J.F.,  Hughes,  M.F.,  Ring,  C.L.,  MacMillan,  D.K.,  Ford,  J.,  Fennell,  T.R.,  Black,  S.R.,  Snyder,  R.W.,  Sipes,  N.S.,  Wetmore,  B.A.,  Westerhout,  J.,  Setzer,  R.W.,  Pearce,  R.G.,  Simmons,  J.E. and  Thomas,  R.S. (2018): Evaluating In Vitro-In Vivo Extrapolation of Toxicokinetics. Toxicol. Sci., 163, 152-169.