INTERACTION OF TRIAZOLAM AND KETOCONAZOLE IN P-GLYCOPROTEIN-DEFICIENT MICE

Transporter-mediated drug-drug interactions: advancement in models, analytical tools, and regulatory perspective

Drug-drug interactions mediated by transporters are a serious clinical concern hence a tremendous amount of work has been done on the characterization of the transporter-mediated proteins in humans and animals. The underlying mechanism for the transporter-mediated drug-drug interaction is the induction or inhibition of the transporter which is involved in the cellular uptake and efflux of drugs. Transporter of the brain, liver, kidney, and intestine are major determinants that alter the absorption, distribution, metabolism, excretion profile of drugs, and considerably influence the pharmacokinetic profile of drugs. As a consequence, transporter proteins may affect the therapeutic activity and safety of drugs. However, mounting evidence suggests that many drugs change the activity and/or expression of the transporter protein. Accordingly, evaluation of drug interaction during the drug development process is an integral part of risk assessment and regulatory requirements. Therefore, this review will highlight the clinical significance of the transporter, their role in disease, possible cause underlying the drug-drug interactions using analytical tools, and update on the regulatory requirement. The recent in-silico approaches which emphasize the advancement in the discovery of drug-drug interactions are also highlighted in this review. Besides, we discuss several endogenous biomarkers that have shown to act as substrates for many transporters, which could be potent determinants to find the drug-drug interactions mediated by transporters. Transporter-mediated drug-drug interactions are taken into consideration in the drug approval process therefore we also provided the extrapolated decision trees from in-vitro to in-vivo, which may trigger the follow-up to clinical studies.

Evaluation system for cell-permeable CYP3A4 inhibitory activity using 1α,25-dihydroxy-vitamin D3-induced intestinal cell lines

We developed an assay system to evaluate the cytochrome P450 (CYP) 3A4-inhibitory activity of compounds, taking account of their cellular permeability, using intestine-derived cell lines pre-treated with the CYP3A4 inducer 1α,25-dihydroxy-vitamin D₃ (250 nM).Ketoconazole (KTZ), saquinavir (SQV), naringin, naringenin (NGE), bergamottin (BG), 6',7'-dihydroxybergamottin (DHBG), epigallocatechin gallate (EGCG), and resveratrol (RES) were evaluated as known CYP3A4 inhibitors. The apparent IC₅₀ (IC_50,app) values of known inhibitors were determined in Caco-2 cells with 10 µM midazolam as a CYP3A4 substrate, and compared with the IC₅₀ values in a human liver microsome assay.SQV and BG with high lipophilicity and good membrane permeability show similar concentrations inside and outside the cells, and consequently IC_50,app and IC₅₀ are similar.KTZ, EGCG, DHBG, NGE, and RES showed a difference between IC₅₀ and IC_50,app. This is considered to result from a difference between the intracellular and extracellular concentrations of the compound, which is likely due to the involvement of efflux and/or influx transporters.This method to evaluate CYP inhibition taking account of membrane permeation should be helpful to assess the potential clinical relevance of drug-drug or drug-food interactions in the gastrointestinal tract.

Kinetic analysis of sequential metabolism of triazolam and its extrapolation to humans using an entero-hepatic two-organ microphysiological system

The microphysiological system (MPS) is a promising tool for predicting drug disposition in humans, although limited information is available on the quantitative assessment of sequential drug metabolism in MPS and its extrapolation to humans. In the present study, we first constructed a mechanism-based pharmacokinetic model for triazolam (TRZ) and its metabolites in the entero-hepatic two-organ MPS, composed of intestinal Caco-2 and hepatic HepaRG cells, and attempted to extrapolate the kinetic information obtained with the MPS to the plasma concentration profiles in humans. In the two-organ MPS and HepaRG single culture systems, TRZ was found to be metabolized into α- and 4-hydroxytriazolam and their respective glucuronides. All these metabolites were almost completely reduced in the presence of a CYP3A inhibitor, itraconazole, confirming sequential phase I and II metabolism. Both pharmacokinetic model-dependent and -independent analyses were performed, providing consistent results regarding the metabolic activity of TRZ: clearance of glucuronidation metabolites in the two-organ MPS was higher than that in the single culture system. The plasma concentration profile of TRZ and its two hydroxy metabolites in humans was quantitatively simulated based on the pharmacokinetic model, by incorporating several scaling factors representing quantitative gaps between the MPS and humans. Thus, the present study provided the first quantitative extrapolation of sequential drug metabolism in humans by combining MPS and pharmacokinetic modeling.

P-Glycoprotein in skin contributes to transdermal absorption of topical corticosteroids

ATP binding cassette transporters, P-glycoprotein (P-gp) and breast cancer resistance protein (BCRP), are expressed in skin, but their involvement in transdermal absorption of clinically used drugs remains unknown. Here, we examined their role in transdermal absorption of corticosteroids. Skin and plasma concentrations of dexamethasone after dermal application were reduced in P-gp and BCRP triple-knockout (Mdr1a/1b/Bcrp^-/-) mice. The skin concentration in Mdr1a/1b/Bcrp^-/- mice was reduced in the dermis, but not in the epidermis, indicating that functional expression of these transporters in skin is compartmentalized. Involvement of these transporters in dermal transport of dexamethasone was also supported by the observation of a higher epidermal concentration in Mdr1a/1b/Bcrp^-/- than wild-type mice during intravenous infusion. Transdermal absorption after dermal application of prednisolone, but not methylprednisolone or ethinyl estradiol, was also lower in Mdr1a/1b/Bcrp^-/- than in wild-type mice. Transport studies in epithelial cell lines transfected with P-gp or BCRP showed that dexamethasone and prednisolone are substrates of P-gp, but are minimally transported by BCRP. Thus, our findings suggest that P-gp is involved in transdermal absorption of at least some corticosteroids in vivo. P-gp might be available as a target for inhibition in order to deliver topically applied drugs and cosmetics in a manner that minimizes systemic exposure.

Astrocyte-mediated regulation of multidrug resistance p-glycoprotein in fetal and neonatal brain endothelial cells: age-dependent effects

Brain endothelial cells (BECs) form a major component of the blood-brain barrier (BBB). In late gestation, these cells express high levels of the multidrug transporter p-glycoprotein (P-gp; encoded by Abcb1), which prevents the passage of an array of endogenous factors and xenobiotics into the fetal brain. P-gp levels in the BECs increase dramatically in late gestation, coincident with astrocyte differentiation. However, the role of astrocytes in modulating P-gp in the developing BBB is unknown. We hypothesized that factors produced by astrocytes positively regulate P-gp in BECs. Astrocytes and BECs were isolated from fetal and postnatal guinea pigs. Levels of Abcb1 mRNA and P-gp were increased in BECs co-cultured with astrocytes compared to BECs in monoculture. Moreover, postnatal astrocytes enhanced P-gp function in fetal BECs but fetal astrocytes had no effect on postnatal BECs. These effects were dependent on secreted proteins with a molecular weight in the range of 3-100 kDa. LC/MS-MS revealed significant differences in proteins secreted by fetal and postnatal astrocytes. We propose that astrocytes are critical modulators of P-gp at the developing BBB. As such, aberrations in astrocyte maturation, observed in neurodevelopmental disorders, will likely decrease P-gp at the BBB. This would allow increased transfer of P-gp endogenous and exogenous substrates into the brain, many of which have neurodevelopmental consequences.

Novel pre-clinical methodologies for pharmacokinetic drug-drug interaction studies: spotlight on "humanized" animal models

Poly-therapy is common due to co-occurrence of several ailments in patients, leading to the elevated possibility of drug-drug interactions (DDI). Pharmacokinetic DDI often accounts for severe adverse drug reactions in patients resulting in withdrawal of drug from the market. Hence, the prediction of DDI is necessary at pre-clinical stage of drug development. Several human tissue and cell line-based in vitro systems are routinely used for screening metabolic and transporter pathways of investigational drugs and for predicting their clinical DDI potentials. However, ample constraints are associated with the in vitro systems and sometimes in vitro-in vivo extrapolation (IVIVE) fail to assess the risk of DDI in clinic. In vitro-in vivo correlation model in animals combined with human in vitro studies may be helpful in better prediction of clinical outcome. Native animal models vary remarkably from humans in drug metabolizing enzymes and transporters, hence, the interpretation of results from animal DDI studies is difficult. With the advent of modern molecular biology and engineering tools, novel pre-clinical animal models, namely, knockout rat/mouse, transgenic rat/mouse with humanized drug metabolizing enzymes and/or transporters and chimeric rat/mouse with humanized liver are developed. These models nearly simulate human-like drug metabolism and help to validate the in vivo relevance of the in vitro human DDI data. This review briefly discusses the application of such novel pre-clinical models for screening various type of DDI along with their advantages and limitations.

Combination antiretroviral therapy (cART) component ritonavir significantly alters docetaxel exposure

PURPOSE

Non-AIDS-defining cancers (NADCs) now exceed rates of AIDS-defining cancers in HIV-positive patients. Treatment of NADCs may be complicated by drug-drug interactions between antiretrovirals and chemotherapy. Docetaxel is a widely used anticancer agent that is primarily metabolized by CYP3A4 enzymes and used to treat NADCs. A preclinical in vivo assessment was performed to gain a better understanding of CYP3-mediated drug-drug interactions between antiretrovirals and docetaxel, as well as to assess any alterations in gene expression with these combinations.

METHODS

Docetaxel (20 mg/kg i.v.) was administered to male FVB mice in the presence and absence of dexamethasone (10 mg/kg p.o. ×4d), efavirenz (25 mg/kg p.o. ×4d), ketoconazole (50 mg/kg p.o.), or ritonavir (12.5 mg/kg p.o.). At various time points, plasma and liver tissue were harvested. Docetaxel concentrations were determined by LC/MS/MS. Pharmacokinetic parameters were calculated. Liver tissue RNA was used to evaluate alterations in Cyp3a11 and Abcb1a gene expression.

RESULTS

Docetaxel exposure was altered by CYP3A4 inhibitors but not by inducers. The CYP3A4 inducers efavirenz and dexamethasone did not have a significant effect on docetaxel exposure (AUC). However, the CYP3A4 inhibitors ritonavir and ketoconazole resulted in a 6.9- and 3.1-fold increase in AUC, respectively. Alterations in gene expression did not account for the altered docetaxel exposure.

CONCLUSIONS

Docetaxel exposure was significantly altered by CYP3A4 inhibitors. Until a definitive clinical trial is performed, docetaxel should be used with caution in patients on a ritonavir-containing antiretroviral regimen or an alternative antineoplastic therapy or antiretroviral regimen should be considered.

Central immune overactivation in the presence of reduced plasma corticosterone contributes to swim stress-induced hyperalgesia

Although it is widely known that immunological, hormonal and nociceptive mechanisms are altered by exposure to repeated stress, the interplaying roles of each function in the development of post-stress hyperalgesia are not completely clear. Thus, we wanted to establish how interleukin 1-beta (IL-1β), corticosterone and microglia interact to contribute in the development of hyperalgesia following repeated forced swim. Rats were subjected to either forced swim, sham swim or non-conditioned. Each group was then treated with minocycline, ketoconazole, or saline. Thermal nociception was measured via the hot plate test, before and after the behavioral conditioning, whereas blood and lumbar spinal cord tissue samples were obtained at the end of the protocol. Serum levels of corticosterone, spinal tissue concentration of IL-1β and spinal OX-42 labeling (microglial marker) were determined. Rats exposed to forced swim stress developed thermal hyperalgesia along with elevated spinal tissue IL-1β, increased OX-42 labeling and relatively diminished serum corticosterone. Pre-treatment with minocycline and ketoconazole prevented the development of thermal hyperalgesia and the increase in IL-1β, without significantly modifying serum corticosterone. These results suggest that the development of forced swim-induced thermal hyperalgesia requires the simultaneous presence of increased spinal IL-1β, microglial activation, and relatively decreased serum corticosterone.

The use of microdialysis techniques in mice to study P-gp function at the blood-brain barrier

An integrated assay system involving dual/triple-probe microdialysis techniques in rats was developed earlier for testing interactions with P-glycoprotein (P-gp) at the blood-brain barrier using quinidine/PSC-833 as a P-gp substrate/inhibitor combination. The aim of the present study was to expand our assay system to mice using microdialysis with simultaneous sampling of blood and brain and to compare the result with a primary mouse brain endothelial cell monolayer (pMBMEC) assay. Brain penetration of quinidine was dose dependent in both anesthetized and awake mice after intraperitoneal drug administration. PSC-833 pretreatment caused a 2.5- to 3.4-fold increase in quinidine levels of brain dialysate samples in anesthetized or awake animals, after single or repeated administration of PSC-833. In pMBMEC, a 2.0- to 2.5-fold efflux ratio was observed in the transcellular transport of quinidine. The P-gp-mediated vectorial transport of quinidine was eliminated by PSC-833. These results indicate that quinidine with PSC-833 is a good probe substrate-reference inhibitor combination for testing drug-drug interactions with P-gp in the in vivo and in vitro mouse systems. With increasing number of humanized transgenic mice, a test system with mouse microdialysis experimentation becomes more important to predict drug-drug interactions in humans.

The concentration of cyclosporine metabolites is significantly lower in kidney transplant recipients with diabetes mellitus

BACKGROUND

Diabetes mellitus is prevalent among kidney transplant recipients. The activity of drug metabolizing enzymes or transporters may be altered by diabetes leading to changes in the concentration of parent drug or metabolites. This study was aimed to characterize the effect of diabetes on the concentration of cyclosporine (CsA) and metabolites.

METHODS

Concentration-time profiles of CsA and metabolites (AM1, AM9, AM4N, AM1c, AM19, and AM1c9) were characterized over a 12-hour dosing interval in 10 nondiabetic and 7 diabetic stable kidney transplant recipients. All patients were male, had nonfunctional CYP3A5*3 genotype, and were on combination therapy with ketoconazole.

RESULTS

The average daily dose (±SD) of CsA was 65 ± 21 and 68 ± 35 mg in nondiabetic and diabetic subjects, respectively (P = 0.550). Cyclosporine metabolites that involved amino acid 1 (AM1, AM19, AM1c) exhibited significantly lower dose-normalized values of area under the concentration-time curve in patients with diabetes. Moreover, during the postabsorption phase (≥3 hours after dose), metabolite-parent concentration ratios for all metabolites, except AM4N, was significantly lower in diabetic patients. The pharmacokinetic parameters of ketoconazole were similar between the 2 groups thus excluding inconsistent ketoconazole exposure as a source of altered CsA metabolism.

CONCLUSIONS

This study indicates that diabetes mellitus significantly affects the concentration of CsA metabolites. Because CsA is eliminated as metabolites via the biliary route, the decrease in the blood concentration of CsA metabolites during postabsorption phase would probably reflect lower hepatic cytochrome P450 3A4 enzyme activity. However, other mechanisms including altered expression of transporters may also play a role. Results of cyclosporine therapeutic drug monitoring in diabetic patients must be interpreted with caution when nonspecific assays are used.

Physiologically based predictions of the impact of inhibition of intestinal and hepatic metabolism on human pharmacokinetics of CYP3A substrates

The first objective of the present study was to predict the pharmacokinetics of selected CYP3A substrates administered at a single oral dose to human. The second objective was to predict pharmacokinetics of the selected drugs in presence of inhibitors of the intestinal and/or hepatic CYP3A activity. We developed a whole-body physiologically based pharmacokinetics (WB-PBPK) model accounting for presystemic elimination of midazolam (MDZ), alprazolam (APZ), triazolam (TRZ), and simvastatin (SMV). The model also accounted for concomitant administration of the above-mentioned drugs with CYP3A inhibitors, namely ketoconazole (KTZ), itraconazole (ITZ), diltiazem (DTZ), saquinavir (SQV), and a furanocoumarin contained in grape-fruit juice (GFJ), namely 6',7'-dihydroxybergamottin (DHB). Model predictions were compared to published clinical data. An uncertainty analysis was performed to account for the variability and uncertainty of model parameters when predicting the model outcomes. We also briefly report on the results of our efforts to develop a global sensitivity analysis and its application to the current WB-PBPK model. Considering the current criterion for a successful prediction, judged satisfied once the clinical data are captured within the 5th and 95th percentiles of the predicted concentration-time profiles, a successful prediction has been obtained for a single oral administration of MDZ and SMV. For APZ and TRZ, however, a slight deviation toward the 95th percentile was observed especially for C(max) but, overall, the in vivo profiles were well captured by the PBPK model. Moreover, the impact of DHB-mediated inhibition on the extent of intestinal pre-systemic elimination of MDZ and SMV has been accurately predicted by the proposed PBPK model. For concomitant administrations of MDZ and ITZ, APZ and KTZ, as well as SMV and DTZ, the in vivo concentration-time profiles were accurately captured by the model. A slight deviation was observed for SMV when coadministered with ITZ, whereas more important deviations have been obtained between the model predictions and in vivo concentration-time profiles of MDZ coadministered with SQV. The same observation was made for TRZ when administered with KTZ. Most of the pharmacokinetic parameters predicted by the PBPK model were successfully predicted within a two-fold error range either in the absence or presence of metabolism-based inhibition. Overall, the present study demonstrated the ability of the PBPK model to predict DDI of CYP3A substrates with promising accuracy.

Comparison of 3 Assay Systems Using a Common Probe Substrate, Calcein AM, for Studying P-gp Using a Selected Set of Compounds

The multidrug resistance protein 1 (MDR1) transporter is the most abundantly investigated adenosine triphosphate (ATP)–Binding Cassette (ABC) transporter protein. Multiple assay systems were developed to study MDR1-mediated transport and possible drug-drug interactions. Yet, as different probe substrates are used in these assays, it is difficult to directly compare the results. In this study, a common probe substrate was applied in 3 assay systems developed to study MDR1: the cellular dye efflux assay, the ATPase assay, and the vesicular transport assay. This probe substrate is calcein acetoxymethyl ester (calcein AM), the acetoxymethyl ester derivative of the fluorescent dye, calcein. Using a common probe allows the investigation of the effect of passive permeability on the result obtained by testing various compounds. In this study, 22 compounds with different logP values were tested in the above-mentioned 3 assay systems. The vesicular transport assay proved most sensitive, detecting 18 of 22 interactions with the protein. The ATPase assay detected 15 interactions, whereas the cellular dye efflux assay was the least sensitive with only 10 hits. A correlation was found between the hydrophobicity of the compound and the ratio of cellular and vesicular transport IC50 values, indicating the effect of passive permeability on the result. Based on hydrophobicity, the current study provides guidelines on applying the most correct tool for studying MDR1 interactions.

Effects of ketoconazole on the biodistribution and metabolism of [11C]loperamide and [11C]N-desmethyl-loperamide in wild-type and P-gp knockout mice

INTRODUCTION

[(11)C]Loperamide and [(11)C]N-desmethyl-loperamide ([(11)C]dLop) have been proposed as radiotracers for imaging brain P-glycoprotein (P-gp) function. A major route of [(11)C]loperamide metabolism is N-demethylation to [(11)C]dLop. We aimed to test whether inhibition of CYP3A4 with ketoconazole might reduce the metabolism of [(11)C]loperamide and [(11)C]dLop in mice, and thereby improve the quality of these radiotracers.

METHODS

Studies were performed in wild-type and P-gp knockout (mdr-1a/b -/-) mice. During each of seven study sessions, one pair of mice, comprising one wild-type and one knockout mouse, was pretreated with ketoconazole (50 mg/kg, ip), while another such pair was left untreated. Mice were sacrificed at 30 min after injection of [(11)C]loperamide or [(11)C]dLop. Whole brain and plasma samples were measured for radioactivity and analyzed with radio-high-performance liquid chromatography.

RESULTS

Ketoconazole increased the plasma concentrations of [(11)C]loperamide and its main radiometabolite, [(11)C]dLop, by about twofold in both wild-type and knockout mice, whereas the most polar radiometabolite was decreased threefold. Furthermore, ketoconazole increased the brain concentrations of [(11)C]loperamide and the radiometabolite [(11)C]dLop by about twofold in knockout mice, and decreased the brain concentrations of the major and most polar radiometabolite in wild-type and knockout mice by 82% and 49%, respectively. In contrast, ketoconazole had no effect on plasma and brain distribution of administered [(11)C]dLop and its radiometabolites in either wild-type or knockout mice, except to increase the low plasma [(11)C]dLop concentration. The least polar radiometabolite of [(11)C]dLop was identified with LC-MS(n) as the N-hydroxymethyl analog of [(11)C]dLop and this also behaved as a P-gp substrate.

CONCLUSION

In this study, ketoconazole (50 mg/kg, ip) proved partially effective for inhibiting the N-demethylation of [(11)C]loperamide in mouse in vivo but had relatively smaller or no effect on [(11)C]dLop.

Effect of micafungin on cytochrome P450 3A4 and multidrug resistance protein 1 activities, and its comparison with azole antifungal drugs

The effects of micafungin on cytochrome P450 3A4 (CYP3A4) metabolic and multidrug resistance protein 1 (MDR1) transport activities were investigated and compared with those of amphotericin B and four azole antifungal drugs (ketoconazole, itraconazole, fluconazole and miconazole). The effects on the metabolic activity of CYP3A4 were examined by measuring nifedipine oxidase activity in human liver microsomes and the effects on MDR1 transport activity were evaluated using [3H]digoxin in MDR1-overexpressing LLC-GA5-COL150 cells. An inhibitory effect on CYP3A4 activity was found for ketoconazole, itraconazole and miconazole, with 50% inhibitory concentrations of 11.7, 32.6 and 74.2 nM, respectively. Fluconazole and micafungin had only slight inhibitory effects and amphotericin B had no effect. The MDR1-mediated transport of [3H]digoxin was inhibited by ketoconazole and itraconazole, and slightly by miconazole. It is suggested that micafungin and amphotericin B would be unlikely to cause drug-drug interactions by inhibition of CYP3A4 and MDR1. A positive correlation between the inhibitory effects on CYP3A4 and MDR1 activities was observed, and the physicochemical mechanisms involved and impact on clinical treatment should be studied further.

Examination of CYP3A and P-glycoprotein-mediated drug-drug interactions using animal models

With the advent of polytherapy for cancer treatment it has become prudent to minimize, as much as possible, the potential for drug-drug interactions (DDI). Toward this end, the metabolic and transporter pathways involved in the disposition of a drug candidate (phenotyping) and potential for inhibition and induction of drug-metabolizing enzymes and transporters are evaluated in vitro. Such in vitro human data can be made available prior to human dosing and enable in vitro to in vivo-based predictions of clinical outcomes. Despite some success, however, in vitro systems are not dynamic and sometimes fail to predict drug-drug interactions for a variety of reasons. In comparison, relatively less effort has been made to evaluate predictions based on data derived from in vivo animal models. This chapter will attempt to summarize different examples from the literature where animal models have been used to predict cytochrome P450 3A (CYP3A)- and P-glycoprotein-based DDI. When employing data from animal models one needs to be aware of species differences in enzyme- and transporter-activity leading to differences in pharmacokinetics, clearance pathways as well as species differences in selectivity and affinity of probe substrates and inhibitors. Because of these differences, in vivo animal studies alone, cannot be predictive of human DDI. Despite these caveats, the information obtained from validated in vivo animal models may prove useful when used in conjunction with in vitro-in vivo extrapolation methods. Such an integrated data set can be used to select drug candidates with a reduced DDI potential.

Drugs as CYP3A probes, inducers, and inhibitors

Human cytochrome P450 (CYP) 3A subfamily members (mainly CYP3A4 and CYP3A5) mediate the metabolism of approximately half all marketed drugs and thus play a critical role in the drug metabolism. A huge number of studies on CYP3A-mediated drug metabolism in humans have demonstrated that CYP3A activity exhibits marked ethnic and individual variability, in part because of altered levels of CYP3A4 expression by various environmental factors and functionally important polymorphisms present in CYP3A5 gene. Accumulating evidence has revealed that CYP3A4 and CYP3A5 have a significant overlapping in their substrate specificity, inducers and inhibitors. Therefore, it is difficult to define their respective contribution to drug metabolism and drug-drug interactions. Furthermore, P-glycoprotein and CYP3A are frequently co-expressed in the same cells and share a large number of substrates and modulators. The disposition of such drugs is thus affected by both metabolism and transport. In this review, we systematically summarized the frequently used CYP3A probe drugs, inducers and inhibitors, and evaluated their current status in drug development and research.

A recursive-partitioning model for blood–brain barrier permeation

A series of bagged recursive partitioning models for log(BB) is presented. Using a LGO-CV, three sets of physical property descriptors are evaluated and found to have Q2 values of 0.51 (CPSA), 0.53 (Ro5x) and 0.53 (MOE). Extrapolating these models to Pfizer chemical space is difficult due to P-glycoprotein (P-gp) mediated efflux. Low correlation coefficients for this test set are improved (R2 = 0.39) when compounds known to be P-gp substrates or statistical extrapolations are removed. The use of simple linear models for specific chemical series is also found to improve the correlation over a limited chemical space.