Quantitative Assessment of the Impact of Fluorine Substitution on P

May 26, 2016 - Development, Groton, Connecticut 06340, United States ...... chamber, t is time, Mr is the mass of compound appearing in the receiver ...
1 downloads 0 Views 5MB Size
Article pubs.acs.org/jmc

Quantitative Assessment of the Impact of Fluorine Substitution on P‑Glycoprotein (P-gp) Mediated Efflux, Permeability, Lipophilicity, and Metabolic Stability Martin Pettersson,*,† Xinjun Hou,† Max Kuhn,‡ Travis T. Wager,† Gregory W. Kauffman,§ and Patrick R. Verhoest† †

Worldwide Medicinal Chemistry, Pfizer Worldwide Research and Development, Cambridge, Massachusetts 02139, United States Research Statistics, Pfizer Worldwide Research and Development, Groton, Connecticut 06340, United States § Computational ADME Group, Department of Pharmacokinetics, Dynamics and Metabolism, Pfizer Worldwide Research and Development, Groton, Connecticut 06340, United States ‡

S Supporting Information *

ABSTRACT: Strategic replacement of one or more hydrogen atoms with fluorine atom(s) is a common tactic to improve potency at a given target and/or to modulate parameters such as metabolic stability and pKa. Molecular weight (MW) is a key parameter in design, and incorporation of fluorine is associated with a disproportionate increase in MW considering the van der Waals radius of fluorine versus hydrogen. Herein we examine a large compound data set to understand the effect of introducing fluorine on the risk of encountering P-glycoprotein mediated efflux (as measured by MDR efflux ratio), passive permeability, lipophilicity, and metabolic stability. Statistical modeling of the MDR ER data demonstrated that an increase in MW as a result of introducing fluorine atoms does not lead to higher risk of Pgp mediated efflux. Fluorine-corrected molecular weight (MWFC), where the molecular weight of fluorine has been subtracted, was found to be a more relevant descriptor.



INTRODUCTION Optimization of physicochemical properties has become an integral part of current medicinal chemistry practice with the goal of reducing attrition due to toxicity and problems associated with absorption, distribution, metabolism, and excretion (ADME).1,2 Consideration of physicochemical properties in medicinal chemistry decision making spans from the very earliest stages of a program such as HTS triage and selection of lead matter to hit-to-lead and lead optimization stages to the final optimization and selection of a clinical candidate. Since the seminal publication by Lipinski in 1997 which established the “rule-of-five”,3 a multitude of propertybased design parameters have emerged to enable evaluation of compound quality. For example, these include ligand efficiency (LE),4 lipophilic efficiency/lipophilic ligand efficiency (LipE/ LLE),5,2 ligand efficiency dependent lipophilicity (LELP),6 lipophilic metabolism efficiency (LipMetE),7 fit quality scaled ligand efficiency (LE_scaled),6 and central nervous system multiple parameter optimization score (CNS MPO).8 A number of analyses have further solidified the central role of physicochemical properties in the successful advancement of drug candidates, and through these efforts, molecular weight (MW) and lipophilicity have emerged as two parameters of particular importance. For example, an analysis by Gleeson © XXXX American Chemical Society

resulted in the guidance of MW < 400 Da and cLogP < 4 (rule of 4/400).9 Johnson proposed a “golden triangle” to describe MW dependence of the logD range that ensures greatest probability for balancing metabolic stability and permeability.10 Although it is now well established that drug molecules with low MW and reduced lipophilicity stand a greater chance of survival, there are several notable exceptions among marketed drugs.11 In addition, some targets are not as readily amenable to development of drug molecules within “ideal physicochemical property space” and may require ligands with a different property profile to achieve adequate target engagement. Aspartyl proteases represent one such challenging target class as exemplified by HIV protease inhibitors, many of which reside in relatively high MW space.12 Additional examples include βsecretase enzyme 1 (BACE1) and γ-secretase, two important targets for Alzheimer’s disease (AD).13 Analyses of the γsecretase modulator (GSM) patent literature clearly indicate that a majority of exemplified compounds are more lipophilic and have considerably higher MW than that of average central nervous system (CNS) drugs.14,15 During our own efforts in this area we found that many of our most interesting lead Received: January 7, 2016

A

DOI: 10.1021/acs.jmedchem.6b00027 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Figure 1. Correlation between MW and P-gp efflux liability (MDR efflux ratio).

Figure 2. Correlation between MW and passive permeability (RRCK).

Figure 3. Examples of strategic use of fluorine: (A) γ-secretase modulators to improve potency and metabolic stability; (B) BACE1 inhibitors to reduce pKa and hERG activity.

molecules had MWs ranging from 400 to 500 Da,16−18 whereas the average MW for CNS drugs on the market is 305 Da.19 High MW is a particular concern for CNS compounds, as this is often associated with poor brain penetration as a result of active efflux mediated by P-glycoprotein (P-gp). A common strategy to assess whether or not a compound may be a substrate for P-glycoprotein (P-gp) mediated efflux is to measure the efflux ratio in a Madin−Darby canine kidney (MDCK) cell line that has been transfected with the human MDR1 gene encoding P-gp.20 In this assay, the rate of transport from the apical to the basolateral side of the cell membrane is compared to drug transport in the opposite direction, and an

MDR efflux ratio (MDR ER) is calculated. An MDR ER > 2.5 is suggestive of an increased risk that a particular compound may be a substrate for P-gp mediated efflux in vivo, whereas MDR ER ≤ 2.5 indicates reduced risk. An examination of the Pfizer compound collection with MW < 600 Da and for which MDR ER data are available (167 314 compounds) clearly shows a strong dependence of MW on P-gp risk as noted previously in the literature.19 As shown in Figure 1, compounds were divided into MW bins of 50 Da and then classified as having MDR ER less than or greater than 2.5. For example, in the cohort of compounds with MWs ranging from 250 to 300 Da (11 327 compounds) only 14.9% were categorized as being substrates B

DOI: 10.1021/acs.jmedchem.6b00027 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

fluorine atoms resulting in an increase in MW from 420 to 492, an increase of 72 Da (non-fluorine heavy atom count is the same for compounds 1 and 2). This is a disproportionate increase in the total MW considering that the van der Waals radius of fluorine is only slightly larger than that of hydrogen (1.47 Å vs 1.20 Å) as has previously been noted in the literature.30 We were interested in examining this phenomenon, as it applies to permeability. Specifically, we sought to determine whether or not a fluorine-derived MW increase leads to an increased risk of P-gp-mediated efflux as determined by MDR ER. We hypothesized that this should not be the case based on our own observations as well as literature reports proposing that scaled MWs be used for compounds containing halogens.30 Another example of strategic use of fluorine was recently published by the Pfizer BACE1 team.26 In this paper, Brodney et al. described the design of compound 4 (Figure 3B) that incorporates three fluorine atoms. Specifically, introduction of fluorine on the methyl substituent on the thioamidine heterocyclic core was hypothesized to reduce hERG activity by reducing pKa. Indeed, this structural modification led to a reduction in pKa from 7.7 to 7.0, which in turn afforded an increase in the hERG IC50 from 2.1 μM to 10.1 μM. Comparison of logD for this matched molecular pair (MMP, compounds 3 and 4) shows an increase of 0.3. Nevertheless, metabolic stability remained high (HLM CLint,app < 8 mL/min/ kg). The authors described additional aliphatic hydrogen to fluorine MMPs indicating a similar trend in pKa, hERG, and logD.

for P-gp efflux (red section) whereas 85.1% were characterized as low risk (blue section). This may then be contrasted to the MW bin containing compounds with MWs between 500 and 550 Da (8928 compounds) where 72.2% of compounds were characterized as being P-gp substrates. Nevertheless, even at high MW there are compounds that have low MDR ER. We became interested in understanding what factors may contribute to a compound having low MDR ER while residing in MW ranges outside those traditionally observed for CNS drugs. This was of particular interest as part of our efforts on the γ-secretase modulator program (vide infra). Passive permeability is another important ADME parameter that is adversely affected by increasing the MW. The passive permeability of a compound can be determined by measuring the rate of flux from the apical to basolateral side of a cell membrane using RRCK cells.21 At the time of the analysis, RRCK data were available for 208 619 compounds with MWs under 600 Da in the Pfizer file. These compounds were then classified as having either reduced permeability (Papp, A→B ≤ 5 × 10−6 cm/s) or high permeability (Papp, A→B > 5 × 10−6 cm/s). In analogy with the plot in Figure 1, Figure 2 displays the data in the form of pie charts where blue sectors represent the percentage of compounds with good permeability and red sectors correspond to the percentage of compounds with reduced permeability within separate MW bins of 50 Da. Consistent with the literature, the data in Figure 2 clearly show that the percentage of compounds that have favorable passive permeability is highest within low MW space and decreases as MW is increased. Incorporation of fluorine into drug molecules is a commonly used tactic to increase potency (LipE) and to optimize the ADME profile (LipMetE); however, it comes at the expense of increased MW and may also impact logD.22 Applications of fluorine in medicinal chemistry include reduction of cyctochrome P450 (CYP450) mediated oxidative metabolism by blocking metabolically labile sites,23 reducing hERG (human ether-a-go-go-related gene) inhibition by lowering the pKa of a basic amine,24−26 and affecting conformational bias.27 It is important to stress that the use of fluorine should be applied strategically to achieve specific ADME or potency objectives, since unproductive introduction of aromatic fluoro substituents may lead to increased lipophilicity and higher metabolic clearance. Paying close attention to the corresponding changes in LipE and LipMetE as a result of replacing a specific hydrogen atom with fluorine is a good approach to evaluate progress. In this manner, incorporation of fluorine played a central role during lead optimization of the Pfizer γ-secretase modulator (GSM) program as highlighted in Figure 3.18 tert-Butyl-substituted GSM library hit 1 (Figure 3A) was identified as an interesting starting point for further optimization, but it had poor metabolic stability and only moderate potency. Incorporation of a trifluoromethyl group allowed for the creation of a tert-butyl replacement with increased metabolic stability that also afforded a significant boost in potency.28,29 Installation of an additional fluorine atom in the para-position of the terminal phenyl ring helped block a site that had previously been shown to be metabolically labile.16 The optimized compound 2 not only had a single digit wholecell IC50 value (Aβ42 IC50 = 6 nM) but also had good metabolic stability (HLM CLint = 12.7 mL/min/kg) resulting in significant increases in LipE and LipMetE while maintaining low MDR ER (1.4). The cost in terms of MW for optimization of 1 into 2 is the net replacement of four hydrogens with four



RESULTS AND DISCUSSION To examine the question of whether or not an increase in MW due to introduction of fluorine atoms leads to a concomitant increase in the risk of a compound being a substrate for P-gp mediated efflux, we set out to analyze data for compounds in the Pfizer corporate compound collection. At the time of the analysis, there were 169 761 unique compounds for which MDR ER data had been acquired. The analysis was limited to compounds with MWs ranging from 300 to 550 Da and fluorine counts of zero to five fluorine atoms. These criteria reduced the compound set to 149 420 unique compounds. The average MDR ER values were then computed over any replicates, and each compound was assigned to one of two bins: MDR ER ≤ 2.5 and MDR ER > 2.5. The distribution of compounds across chemotype (acids, bases, neutrals, and zwitterions) as well as physicochemical property space (fraction sp3, cLogD, and TPSA) is described in detail in the Supporting Information (Figures SI-1−SI-5). 81% of the compounds in the data set are neutral, and the distribution is similar across the MW range of the analysis. Likewise, the distribution across fraction sp3 remains similar across MW space. Each compound in the collection was further characterized according to fluorine count ranging from zero to five, where fluorine count refers to the total number of fluorine atoms in a given molecule. This includes fluorine atoms in different environments such as in compounds 2 and 4 where both “aliphatic fluorines” and “arylfluorines” contribute to total fluorine count. A characterization of the various types of fluoro substituents of all compounds across the MW range of the data set can be found in the Supporting Information (see Figure SI6). The analysis shows that a majority of fluorine atoms reside in the aryl-F or aryl-CHmFn (for example, aryl-CF3 or arylC

DOI: 10.1021/acs.jmedchem.6b00027 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Figure 4. Probability plot indicating the likelihood of compounds having MDR ER > 2.5 at a given MW for compounds with fluorine counts ranging from zero to five.

Figure 5. Probability plot indicating the likelihood of compounds having MDR ER > 2.5 at a given fluorine-corrected MW (MWFC) for compounds with fluorine counts ranging from zero to five.

Figure 4 displays the probability of a compound having MDR ER > 2.5 at a given MW for compounds segregated into fluorine counts ranging from zero to five. In accordance with the trend observed in Figure 1, there is a clear relationship between MW and the likelihood of a compound having MDR ER > 2.5. As expected, this relationship holds true regardless of the number of fluorine atoms in the molecule. However, an interesting observation emerged from this analysis: At any given MW, the probability of a compound having MDR ER > 2.5 is strongly dependent on the number of fluorine atoms that are present in the molecule. For example, at a fixed MW of 400, compounds containing no fluorine atoms have a probability of about 0.5 of having an MDR ER > 2.5 (Figure 4). If the MW is kept constant at 400 but the number of fluorine atoms is increased, the MDR ER liability decreases such that with five fluorine atoms, the likelihood of MDR ER > 2.5 is reduced to

CHF2), and this remains constant over the MW range of the data set. To quantify the relationship between the probability of MDR ER > 2.5 and molecular descriptors, logistic regression was used, specifically within the generalized additive model framework.31 This approach provides an opportunity to easily incorporate nonlinearities into the relationship between each descriptor and the odds of P-gp liability without increasing model complexity or losing the ability to conduct formal statistical inference (when required). Spline basis functions were used in the model and the smoothness was determined using 10-fold cross-validation.32 Additionally, this model can produce measures of uncertainty to allow the noise in the relationships to be visualized via 95% confidence bands for model probabilities. D

DOI: 10.1021/acs.jmedchem.6b00027 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Figure 6. Contour plot indicating probability of MDR ER > 2.5 as a function of logD and MW for fluorine counts ranging from zero to five.

Figure 7. Contour plot indicating probability of MDR ER > 2.5 as a function of pKa and MW for fluorine counts ranging from zero to five.

MDR ER liability at a given MWFC is independent of fluorine count. In examination of the data in Figure 5 more closely, several interesting observations emerge. For example, for compounds with MWFC = 400, the fraction of compounds having MDR ER > 2.5 is about 0.5. This means that compounds with MWFC = 400 and no fluorine atoms have the same MDR ER liability as compounds with MWFC = 400 and five fluorine atoms. That corresponds to an increase in the total MW by 90 Da. This analysis therefore indicates that an increase in total MW due to addition of fluorine atoms does not increase the MDR ER liability. This observation holds true across the entire MW spectrum examined in the analysis, thus supporting the original hypothesis. This finding makes intuitive sense given that MW can be thought of as a surrogate for the size, or the molecular volume, which is a key factor in molecular recognition. Since the van der Waals radius of a fluorine atom is only slightly larger than that of a hydrogen atom (1.47 Å vs 1.20 Å), the molecular volume does not increase to the same extent as MW when a hydrogen atom is replaced with a fluorine atom. Therefore, MW is not a good surrogate for molecular volume for drug molecules containing fluorine atoms.30 Figures 4 and 5 clearly demonstrate that MDR ER liability decreases as the number of fluorine atoms increases and makes up a larger part of the total MW, which is consistent with the original hypothesis. However, in addition to the consideration of molecular volume, there may be other factors that contribute to this effect. For example, incorporation of fluorine in close proximity to a basic nitrogen atom may lead to a reduction in

0.25. Again, this trend holds true across the entire range of MWs that were examined giving rise to six separate curves, one for each fluorine count. The data therefore suggest that as the MW contribution of non-fluorine atoms decreases, the probability that a compound will have MDR ER > 2.5 decreases as well, which is in accordance with the original hypothesis. To further examine these trends, we sought to evaluate the relationship between MDR ER and the MW component derived from non-fluorine atoms. To facilitate this analysis, we proposed a molecular descriptor termed “fluorine-corrected molecular weight” (MWFC), which is defined as shown in eq 1.33 MWFC = (total MW) − (MW derived from fluorine atoms) (1)

As an example, GSM lead 2 (Figure 3A) has a total MW of 492, but the fluorine corrected MW (MWFC), where the MW of four fluorine atoms has been subtracted, is 416. The MWFC values were then calculated for all compounds in the data set, and the statistical model described above was applied to examine the relationship between MWFC and MDR ER. As shown in Figure 5, there is a consistent dependence of MWFC on the probability of compounds having MDR ER > 2.5. In fact, by use of MWFC instead of MW, the curves for the separate fluorine counts completely overlap such that MWFC can be used to describe MDR ER liability without the need to distinguish compounds based on the number of fluorine atoms. In other words, the E

DOI: 10.1021/acs.jmedchem.6b00027 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Figure 8. Plot of molecular volumes calculated from Corina-derived conformations vs molecular volumes obtained from the corresponding X-ray structures.

pKa,34 which in turn has been shown to correlate with reduced MDR ER.8 It is also plausible that changes in lipophilicity as a result of introduction of fluorine may impact MDR ER, as it has previously been demonstrated that passive permeability is correlated with logD.10 It is important to note that not all fluorine for hydrogen substitutions lead to increased lipophilicity. Although introduction of aromatic fluorine atoms generally increases lipophilicity, aliphatic fluorine atoms may decrease lipophilicity as a result of increased dipole moment (vide infra).35,36 Nevertheless, we were interested in re-examining the data set by normalizing for potential changes in lipophilicity. Toward this end, measured logD 37 was incorporated into the statistical model as shown in Figure 6, and the risk of encountering P-gp mediated efflux is displayed as contour maps with dark blue indicating increased probability of MDR ER > 2.5 whereas lighter color corresponds to reduced probability of MDR ER > 2.5. logD is plotted on the Y-axis and total MW on the X-axis, and the plots are trellised according to fluorine count. Using this visualization, one can examine the effect of increased fluorine count on MDR ER at a constant logD. For example, focusing on the MDR ER probability associated with logD = 3 and total MW = 400 as indicated by the cross-hairs, it is evident that as fluorine count increases, the MDR ER liability decreases, which is consistent with the observations discussed above. This trend holds true across the MW and logD spectrum and therefore indicates that the reduction in MDR ER liability with increased fluorine count is independent of lipophilicity. We next sought to evaluate the MDR ER data set while normalizing for potential changes in pKa. As shown in Figure 7, pKa was incorporated into the statistical model, and the probability of encountering MDR ER > 2.5 is again displayed as

contour maps with darker color indicating higher probability and lighter color indicating lower probability of MDR ER > 2.5. pKa is plotted on the Y-axis, total MW is shown on the X-axis, and the plots are then trellised according to fluorine count from zero to five fluorine atoms. This visualization allows for evaluation the impact of increased fluorine count on MDR ER liability while keeping pKa constant. For example, focusing on pKa of 5 and MW of 400 as shown by the cross-hairs demonstrates that as fluorine count increases (i.e., the MW of non-fluorine atoms decreases), the probability of encountering MDR ER > 2.5 decreases. This trend holds true across the MW and pKa space and therefore indicates that the reduction in MDR ER liability with increased fluorine count is independent of changes in pKa, thus supporting the original hypothesis. Having confirmed the original hypothesis, we were next interested in evaluating the relative effect of molecular volume vs MW on MDR ER liability. Given that MW may be thought of as a surrogate for molecular size and that the size of fluorine is relative small versus other heavy atoms, it follows that molecular volume should be a more relevant descriptor and that MDR ER liability should be more strongly correlated to molecular volume as compared to total MW. While several molecular volume algorithms exist,38 we sought a method that could be readily accessible for medicinal chemists and therefore adapted the “molecular volume” method of Pipeline Pilot.39 Since molecular volume is a property depending on the threedimensional conformational structure of a molecule, we evaluated the conformational dependency of Pipeline Pilot’s algorithm using a set of small molecules extracted from crystal structure database of protein−ligand complex in PDB Ligand Expo.40 Of the approximate 13K unique small organic molecules with MW 250−800 (see Supporting Information), F

DOI: 10.1021/acs.jmedchem.6b00027 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

the entire data set of 149 420 compounds irrespective of fluorine count. The resultant visualization enables examination of the effect of increasing MW while holding molecular volume constant and vice versa. Interestingly, if MW is increased while molecular volume is held constant, there is only a marginal gradient of increase in MDR ER liability (Figure 9, arrow A). On the contrary, if molecular volume is increased while holding MW constant, there is a considerable gradient of increase in MDR ER liability (Figure 9, arrow B). The data in Figure 9 clearly demonstrate that changes in MW that do not lead to a corresponding change in molecular volume do not significantly impact MDR ER liability. This finding has potentially interesting implication for drug design as it suggests that molecular volume may be a more relevant medicinal chemistry design parameter than the ubiquitously used MW. Furthermore, implementing a work-flow whereby molecular volume is calculated using Corina-based 3D conformations may then provide a straightforward and reliable approach to generate molecular volume data at the design stage. Having confirmed that an increase in MW due to addition of fluorine atoms does not lead to an increased risk of encountering P-gp-mediated efflux and that molecular volume is a more relevant descriptor than MW, we next hypothesized that these findings could be generalized to passive permeability. To examine this idea, we analyzed passive permeability data (RRCK Papp,A→B),21 which was available for 186 490 compounds in the Pfizer file that had MW between 300 and 550 Da and zero to five fluorine atoms. Each compound was assigned as having either good permeability (RRCK Papp,A→B > 5.0 × 10−6 cm/s) or reduced permeability (RRCK Papp,A→B ≤ 5.0 × 10−6 cm/s). It has already been shown that as MW increases, passive permeability generally decreases,10 and this is evident from our data set as well. As shown in Figure 10, increasing the MW regardless of fluorine count reduces the likelihood of having good passive permeability (RRCK Papp,A→B > 5.0 × 10−6 cm/s). However, the probability is dependent on the number of fluorine atoms in the molecule giving rise to six separate lines (one for each fluorine count from zero to five). The data indicate that for any given MW, compounds with increased fluorine count have increased probability of having RRCK

no significant variations were observed when computed with either 3D conformation from the original crystal structure, Corina generated structures (Figure 8),41 or OpenEye OMEGA42 generated structures (see Supporting Information). Pipeline Pilot’s 3D coordinate component also created a reasonable correlation but with a slightly larger variation as compared to Corina or OpenEye OMEGA conformations. In the following analysis we use the Corina generated conformations to compute molecular volume. With a reliable method in hand for calculating molecular volume, we next sought to compare the relative impact of molecular volume vs MW on MDR ER. Figure 9 shows a

Figure 9. Contour plot indicating probability of MDR ER > 2.5 as a function of changes in MW and molecular volume.

probability contour map with respect to molecular volume vs MW where dark blue indicates high probability of MDR ER > 2.5 and lighter color corresponds to reduced probability of MDR ER > 2.5. White regions refer to areas in the descriptor space where data were not available. The analysis incorporates

Figure 10. Probability plot indicating the likelihood of compounds having RRCK Papp,A→B > 5.0 × 10−6 cm/s at a given MW for compounds with fluorine counts ranging from zero to five. G

DOI: 10.1021/acs.jmedchem.6b00027 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Figure 11. Probability plot indicating the likelihood of compounds having RRCK Papp,A→B > 5.0 × 10−6 cm/s at a at a given fluorine-corrected MW (MWFC) for compounds with fluorine counts ranging from zero to five.

Figure 12. Contour plot indicating probability of RRCK Papp,A→B > 5.0 × 10−6 cm/s as a function of logD and MW for fluorine counts ranging from zero to five.

Papp,A→B > 5.0 × 10−6 cm/s. In analogy to MDR ER analysis, the corollary involves modeling MWFC (Figure 11) where the MW of the fluorine atoms has been subtracted. In this case, the six lines collapse into a single probability curve suggesting that it is the MW component derived from non-fluorine atoms that best correlates with permeability. It has also been established that passive permability is dependent on lipophilicity.10 Figure 12 shows permeability as contour plots, and the data are separated by fluorine count with MW on the X-axis and measured logD on the Y-axis. The data confirm a strong dependence of RRCK on lipophilicity, but a closer inspection of Figure 12 led to several interesting insights: The logD range associated with high probability of having good passive permeability (RRCK Papp,A→B > 5.0 × 10−6 cm/s) decreases as MW increases. There are optimal values for logD at any given fluorine count and MW resulting in U-shaped curves along the logD axis. More importantly, the center of the optimal range remains constant at a logD of about 2.7 (the horizontal line across the six panels in Figure 12) regardless of fluorine count. This finding supports the notion that the improvement in passive permeability as fluorine makes up a larger fraction of MW cannot be attributed to a change in lipophilicity.

Having established that hydrogen-to-fluorine substitution and the associated increase in MW generally does not increase the risk of P-gp mediated efflux or lead to reduced passive permeability, we next sought to examine the associated “cost” in terms of a potential increase in lipophilicity. In a detailed analysis of matched molecular pairs (MMPs), Xing et al. recently showed that the Ar-OCH3 to Ar-OCF3 transform is associated with a significant increase in lipophilicity.22b However, it has also been noted that lipophilicity does not always increase when substituting fluorine for hydrogen. In fact, a recent study by Carreira and co-workers showed that certain aliphatic hydrogen-to-fluorine substitutions may actually lead to a decrease in lipophilicity.36 We were interested in broadly evaluating the effect of fluorine substitution on lipophilicity and therefore selected 11 common hydrogen-to-fluorine transforms for further examination. For each transform we identified all MMPs in the Pfizer compounds collection limiting the set to exact pairs differing only by replacing one or more hydrogen atoms with fluorine atom(s).43 The data set was further restricted to MMPs where measured logD37 or human liver microsomal stability (HLM CLint,app) was available for both compounds in a given pair.44,45 Finally, the change in measured logD (Δ logD), the change in microsomal stability (Δ logH

DOI: 10.1021/acs.jmedchem.6b00027 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

Figure 13. The change in logD, log(HLM CLint,app), and LipMetE is displayed as box plots for 11 common hydrogen-to-fluorine MMP transforms. The number of pairs, median change, and the number of outliers are shown for the three parameters for each transform.

Clearance generally increases as lipophilicity increases, so one might have expected that clearance should increase across the various transforms. However, as shown in Figure 13, Δ log(HLM) remained relatively unchanged. This may be explained by the fact that an increase in lipophilicity can be counteracted by the ability of fluorine to block potentially metabolically labile sites. This effect is most pronounced in the case of phenyl OCH3 to phenyl OCF3 substitution. While this transform is associated with greatest increase in logD, it also eliminates a potential metabolically labile site by replacing the methoxy substituent with a trifluoromethoxy group in addition to reducing the electron density of the phenyl ring. Accordingly, clearance does not increase as one might have expected based on the change in lipophilicity alone. The relationship between metabolic stability and lipophilicity can be assessed by using LipMetE.7 An increase in lipophilicity while avoiding a concomitant reduction in metabolic stability results in increased LipMetE. In other words, introduction of fluorine leads to an increase in LipMetE for a majority of the transforms in Figure 13. In particular, the OCH3 to OCF3 transform is associated

(HLM)), and the change in LipMetE (ΔLipMetE) were then calculated for each MMP, and the data are displayed in Figure 13 as a box plots for each transform. As shown in Figure 13, there is a clear difference in Δ logD across various transforms, which is consistent with previous reports. For example, replacing a phenyl CH3 with phenyl CHF2 leads to a reduction in logD (median Δ logD = −0.25 for 15 MMPs), whereas phenyl OCH3 to phenyl OCF3 resulted in the greatest increase in logD (median Δ logD = 1.02 for 115 MMPs).22b In this data set, replacing an aliphatic CH3 with an aliphatic CF3 resulted in a moderate increase in logD (median Δ logD = 0.43 for 332 MMPs), whereas the Δ logD corresponding to replacement of an aliphatic CH3 with aliphatic CHF2 remained unchanged (101 MMPs). As is evident from the data set, Δ logD can differ significantly within a given transform presumably as a result of how fluorine substitution affects the local molecular environment such as molecular conformation, pKa, etc.46 After having established Δ logD values for each transform, we then analyzed the corresponding changes in metabolic stability. I

DOI: 10.1021/acs.jmedchem.6b00027 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

without causing large disruptions in the steric environment. However, in spite of the small size, introduction of fluorine comes at the expense of a large increase in MW. We have analyzed MDR ER data for 149 420 compounds containing zero to five fluorine atoms to determine whether or not an increase in MW due to addition of fluorine leads to a corresponding increase in MDR ER liability. This analysis clearly demonstrates that increased MW due to introduction of fluorine does not lead to increased MDR ER liability. As a corollary, the probability of encountering high MDR ER at a given MW differs considerably depending on the number of fluorine atoms in a molecule. This may be explained by the fact that MW is not always a good surrogate for molecular size, especially not in the case of fluorinated small molecules. In contrast to MW, using “fluorine corrected MW” (MWFC), in which the MW contribution from fluorine atoms has been subtracted, provides superior correlation with MDR ER liability regardless of fluorine count. The analysis inspired a broader comparison of the impact of molecular volume vs MW on MDR ER liability, and these results demonstrate that molecular volume is a better determinant of MDR ER liability than total MW. We next expanded the analysis to include passive permeability as measured by the RRCK assay and observed the same trend as with the MDR data. While higher MW generally leads to reduced passive permeability, we found that increased MW due to addition of fluorine atoms does not lead to increased probability of impaired passive permeability. Finally, matched molecular pairs of common hydrogen to fluorine transforms were analyzed with respect to measured logD and the corresponding change in HLM stability. While several hydrogen to fluorine transforms result in increased lipophilicity, some transforms are lipophilicity neutral and others even lead to a reduction in lipophilicity, which is in accordance with previously reported data. HLM stability remains relatively unchanged even for transforms associated with increased lipophilicity, which may be due to blockage of metabolically labile sites and is reflected in increased LipMetE. The findings described herein have broader implication for drug design in general, suggesting that molecular volume may be a more relevant design parameter than MW as a measure of molecular size. While we have highlighted a straightforward and reliable method for calculating molecular volume that takes 3D conformation into account, our data indicate that fluorine corrected MW (MWFC) may serve as a valuable and easy-to-use surrogate.

with the highest median increase in LipMetE (0.84). With that said, it is important to note that Figure 13 is aggregating data over a large number of MMPs and that for specific pairs, one should only expect an improvement in metabolic stability if fluorine is introduced strategically to mitigate a metabolic liability. Finally, we were interested in comparing the effect of fluorine versus chlorine on lipophilicity. As in the case of fluorinecontaining compounds, there is indeed a separation in MDR ER liability for compounds of different chlorine count at any given MW, which is consistent with the notion that scaled MW be considered for halogen containing compounds (see Supporting Information, Figure SI-10).30 However, as shown in Figure 14, introduction of a chlorine atom on a phenyl ring

Figure 14. Change in logD is displayed as box plots for two relevant hydrogen-to-chlorine MMP transforms. The number of pairs, median change, and the number of outliers are indicated for each transform.

(Ar-Cl) or heteroaryl ring (Het-Cl) is associated with a higher “cost” in terms of increased lipophilicity with median increases in logD of 0.54 and 0.67 for Ar-Cl and Het-Cl, respectively. In addition, chlorine is not as versatile as fluorine because many chlorine-containing substructures such as aliphatic chlorines may be reactive. Clearly, fluorine is uniquely suited for strategic use in medicinal chemistry for improving potency, tuning physicochemical properties, and optimizing ADME parameters.





CONCLUSION Optimization of physicochemical properties has become a key component of modern drug design to ensure alignment of potency and ADME and to reduce the risk of attrition due to toxicity. CNS targets impose further restrictions on “druggable property space” owing to the added challenge of crossing the blood−brain barrier and the known correlation between MW and P-gp mediated efflux. Introduction of fluorine atoms into drug molecules is a common tactic not only to improve potency but also to optimize ADME parameters such as improving metabolic stability. When used strategically, this approach can result in dramatic improvements in compound quality as gauged by design metrics such as LipE and LipMetE. Furthermore, the small size of fluorine allows it to be readily incorporated into a molecule to tune electronic properties

EXPERIMENTAL SECTION

MDR1 Efflux Ratio Assay. A 96-well Transwell assay method similar to that described previously was used.20 On day 1, human MDCKII-MDR1 cells were seeded at a cell density range of (2.1−3.0) × 105 cells/mL in their corresponding media on 96-well inserts. The cells were cultured at 37 °C, 90% humidity, and 5% CO2, and the assays were performed on day 5. Transwell Data Analysis. The methods similar to that described previously20 were used to determine compound apparent permeability (Papp) values for Transwell studies. The Papp was calculated using eq 2:

Papp =

dM r 1 Area·C D(0) dt

(2)

where Area is the surface area of the cell monolayer (0.0804 cm2), CD(0) is the initial concentration of compound applied to the donor chamber, t is time, Mr is the mass of compound appearing in the receiver compartment as a function of time, and dMr/dt is the flux of J

DOI: 10.1021/acs.jmedchem.6b00027 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

in vivo toxicological outcomes. Bioorg. Med. Chem. Lett. 2008, 18, 4872−4875. (2) (a) Leeson, P. D.; Springthorpe, B. The influence of drug-like concepts on decision-making in medicinal chemistry. Nat. Rev. Drug Discovery 2007, 6, 881−890. (b) Leeson, P. D.; Young, R. J. Molecular property design: Does everyone get it? ACS Med. Chem. Lett. 2015, 6, 722−725. (3) Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv. Drug Delivery Rev. 1997, 23, 3−25. (4) Hopkins, A. L.; Groom, C. R.; Alex, A. Ligand efficiency: a useful metric for lead selection. Drug Discovery Today 2004, 9, 430−431. (5) Ryckmans, T.; Edwards, M. P.; Horne, V. A.; Correia, A. M.; Owen, D. R.; Thompson, L. R.; Tran, I.; Tutt, M. F.; Young, T. Rapid assessment of a novel series of selective CB2 agonists using parallel synthesis protocols: A Lipophilic Efficiency (LipE) analysis. Bioorg. Med. Chem. Lett. 2009, 19, 4406−4409. (6) Reynolds, C. H.; Tounge, B. A.; Bembenek, S. D. Ligand binding efficiency: trends, physical basis, and implications. J. Med. Chem. 2008, 51, 2432−2438. (7) Stepan, A. F.; Kauffman, G. W.; Keefer, C. E.; Verhoest, P. R.; Edwards, M. Evaluating the differences in cycloalkyl ether metabolism using the design parameter "lipophilic metabolism efficiency" (LipMetE) and a matched molecular pairs analysis. J. Med. Chem. 2013, 56, 6985−6990. (8) Wager, T. T.; Hou, X.; Verhoest, P. R.; Villalobos, A. Moving beyond rules: The development of a central nervous system multiparameter optimization (CNS MPO) approach to enable alignment of druglike properties. ACS Chem. Neurosci. 2010, 1, 435−449. (9) Gleeson, M. P. Generation of a set of simple, interpretable ADMET rules of thumb. J. Med. Chem. 2008, 51, 817−834. (10) Johnson, T. W.; Dress, K. R.; Edwards, M. Using the Golden Triangle to optimize clearance and oral absorption. Bioorg. Med. Chem. Lett. 2009, 19, 5560−5564. (11) Rosenquist, Å; Samuelsson, B.; Johansson, P.-O.; Cummings, M. D.; Lenz, O.; Raboisson, P.; Simmen, K.; Vendeville, S.; de Kock, H.; Nilsson, M.; Horvath, A.; Kalmeijer, R.; de la Rosa, G.; BeumontMauviel, M. Discovery and development of simeprevir (TMC435), a HCV NS3/4A protease inhibitor. J. Med. Chem. 2014, 57, 1673−1693. (12) Ghosh, A. K.; Osswald, H. L.; Prato, G. Recent progress in the development of HIV-1 protease inhibitors for the treatment of HIV/ AIDS J. Med. Chem. 2016, DOI: 10.1021/acs.jmedchem.5b01697. (13) De Strooper, B.; Vassar, R.; Golde, T. The secretases: enzymes with therapeutic potential in Alzheimer’s disease. Nat. Rev. Neurol. 2010, 6, 99−107. (14) Pettersson, M.; Stepan, A. F.; Kauffman, G. W.; Johnson, D. S. Novel γ-secretase modulators for the treatment of Alzheimer’s disease: a review focusing on patents from 2010 to 2012. Expert Opin. Ther. Pat. 2013, 23, 1349−1366. (15) Oehlrich, D.; Berthelot, D. J.-C.; Gijsen, H. J. M. γ-Secretase modulators as potential disease modifying anti-Alzheimer’s drugs. J. Med. Chem. 2011, 54, 669−698. (16) Pettersson, M.; Johnson, D. S.; Subramanyam, C.; Bales, K. R.; am Ende, C. W.; Fish, B. A.; Green, M. E.; Kauffman, G. W.; Mullins, P. B.; Navaratnam, T.; Sakya, S. M.; Stiff, C. M.; Tran, T. P.; Xie, L.; Zhang, L.; Pustilnik, L. R.; Vetelino, B. C.; Wood, K. M.; Pozdnyakov, N.; Verhoest, P. R.; O’Donnell, C. J. Design, synthesis, and pharmacological evaluation of a novel series of pyridopyrazine-1,6dione γ-secretase modulators. J. Med. Chem. 2014, 57, 1046−1062. (17) Pettersson, M.; Johnson, D. S.; Humphrey, J. M.; Am Ende, C. W.; Evrard, E.; Efremov, I.; Kauffman, G. W.; Stepan, A. F.; Stiff, C. M.; Xie, L.; Bales, K. R.; Hajos-Korcsok, E.; Murrey, H. E.; Pustilnik, L. R.; Steyn, S. J.; Wood, K. M.; Verhoest, P. R. Discovery of indolederived pyridopyrazine-1,6-dione γ-secretase modulators that target presenilin. Bioorg. Med. Chem. Lett. 2015, 25, 908−913. (18) Pettersson, M.; Johnson, D. S.; Humphrey, J. M.; Butler, T. W.; Am Ende, C. W.; Fish, B. A.; Green, M. E.; Kauffman, G. W.; Mullins,

the compound across the cell monolayer. The efflux ratio (ER) was calculated using eq 3: ER =

Papp(B−A) Papp(A−B)

(3)

where A−B and B−A denote the transport direction in which Papp was determined. RRCK, HLM, logD Assays. Passive permeability (RRCK),21 human liver microsome (HLM) stability,44,45 and lipophilicity (shake flask logD)37 were measured using previously published procedures.22b LipMetE Analysis. Lipophilic metabolism efficiency (LipMetE) was calculated for each compound according to the following equation: LipMetE = logD − log(CLint,u).7 Unbound intrinsic clearance (CLint,u) was derived from the measured human liver microsomal intrinsic apparent clearance (CLint,app)44,45 and calculated human liver microsomal binding values.



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jmedchem.6b00027. Detailed characterization of the compound collection with respect to chemotype and physicochemical properties; comparison of methods for calculating molecular volume; analysis of the effect of introducing chlorine on MDR ER (PDF)



AUTHOR INFORMATION

Corresponding Author

*Phone: (617) 395-0705. E-mail: martin.pettersson@pfizer. com. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS We thank Christopher Keefer for his contributions to the matched molecular pairs analysis and Brajesh Raj for contributions to characterizing the fluorine environment. We also thank the Pfizer ADME technology group for generating the in vitro pharmacokinetic data, and Bo Feng for contributions to the MDR ER assay.



ABBREVIATIONS USED AD, Alzheimer’s disease; ADME, absorption, distribution, metabolism, and excretion; BACE, β-secretase converting enzyme; CLint,app, apparent intrinsic clearance; CNS MPO, central nervous system multiparameter optimization; CYP, cytochrome P450; ER, efflux ratio; GSM, γ-secretase modulator; hERG, human ether-a-go-go-related gene; HLM, human liver microsome; LE, ligand efficiency; LipE, lipophilic efficiency; LLE, lipophilic ligand efficiency; LipMetE, lipophilic metabolism efficiency; HTS, high-throughput screen; MDCK, Madin−Darby canine kidney; MDR1, multidrug resistance protein (P-glycoprotein, P-gp); MW, molecular weight; MWFC, fluorine-corrected molecular weight; RRCK, Ralph Russ canine kidney



REFERENCES

(1) Hughes, J. D.; Blagg, J.; Price, D. A.; Bailey, S.; DeCrescenzo, G. A.; Devraj, R. V.; Ellsworth, E.; Fobian, Y. M.; Gibbs, M. E.; Gilles, R. W.; Greene, N.; Huang, E.; Krieger-Burke, T.; Loesel, J.; Wager, T.; Whiteley, L.; Zhang, Y. Physiochemical drug properties associated with K

DOI: 10.1021/acs.jmedchem.6b00027 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

P. B.; O’Donnell, C. J.; Stepan, A. F.; Stiff, C. M.; Subramanyam, C.; Tran, T. P.; Vetelino, B. C.; Yang, E.; Xie, L.; Bales, K. R.; Pustilnik, L. R.; Steyn, S. J.; Wood, K. M.; Verhoest, P. R. Design of pyridopyrazine-1,6-dione γ-secretase modulators that align potency, MDR efflux ratio, and metabolic stability. ACS Med. Chem. Lett. 2015, 6, 596−601. (19) Wager, T. T.; Chandrasekaran, R. Y.; Hou, X.; Troutman, M. D.; Verhoest, P. R.; Villalobos, A.; Will, Y. Defining desirable central nervous system drug space through the alignment of molecular properties, in vitro ADME, and safety attributes. ACS Chem. Neurosci. 2010, 1, 420−434. (20) Feng, B.; Mills, J. B.; Davidson, R. E.; Mireles, R. J.; Janiszewski, J. S.; Troutman, M. D.; de Morais, S. M. In vitro P-glycoprotein assays to predict the in vivo interactions of P-glycoprotein with drugs in the central nervous system. Drug Metab. Dispos. 2008, 36, 268−275. (21) Callegari, E.; Malhotra, B.; Bungay, P. J.; Webster, R.; Fenner, K. S.; Kempshall, S.; LaPerle, J. L.; Michel, M. C.; Kay, G. G. A comprehensive non-clinical evaluation of the CNS penetration potential of antimuscarinic agents for the treatment of overactive bladder. Br. J. Clin. Pharmacol. 2011, 72, 235−246. (22) (a) Gillis, E. P.; Eastman, K. J.; Hill, M. D.; Donnelly, D. J.; Meanwell, N. A. Applications of fluorine in medicinal chemistry. J. Med. Chem. 2015, 58, 8315−8359. (b) Xing, L.; Blakemore, D. C.; Narayanan, A.; Unwalla, R.; Lovering, F.; Denny, R. A.; Zhou, H.; Bunnage, M. E. Fluorine in drug design: a case study with fluoroanisoles. ChemMedChem 2015, 10, 715−726. (c) Hagmann, W. K. The many roles for fluorine in medicinal chemistry. J. Med. Chem. 2008, 51, 4359−4369. (d) Purser, S.; Moore, P. R.; Swallow, S.; Gouverneur, V. Fluorine in medicinal chemistry. Chem. Soc. Rev. 2008, 37, 320−330. (23) (a) Miller, M. M.; Liu, Y.; Jiang, J.; Johnson, J. A.; Kamau, M.; Nirschl, D. S.; Wang, Y.; Harikrishnan, L.; Taylor, D. S.; Chen, A. Y. A.; Yin, X.; Seethala, R.; Peterson, T. L.; Zvyaga, T.; Zhang, J.; Huang, C. S.; Wexler, R. R.; Poss, M. A.; Lawrence, R. M.; Adam, L. P.; Salvati, M. E. Identification of a potent and metabolically stable series of fluorinated diphenylpyridylethanamine-based cholesteryl ester transfer protein inhibitors. Bioorg. Med. Chem. Lett. 2012, 22, 6503−6508. (b) Zhu, Y.; Olson, S. H.; Graham, D.; Patel, G.; HermanowskiVosatka, A.; Mundt, S.; Shah, K.; Springer, M.; Thieringer, R.; Wright, S.; Xiao, J.; Zokian, H.; Dragovic, J.; Balkovec, J. M. Phenylcyclobutyl triazoles as selective inhibitors of 11β-hydroxysteroid dehydrogenase type I. Bioorg. Med. Chem. Lett. 2008, 18, 3412−3416. (24) Hameed P, S.; Patil, V.; Solapure, S.; Sharma, U.; Madhavapeddi, P.; Raichurkar, A.; Chinnapattu, M.; Manjrekar, P.; Shanbhag, G.; Puttur, J.; Shinde, V.; Menasinakai, S.; Rudrapatana, S.; Achar, V.; Awasthy, D.; Nandishaiah, R.; Humnabadkar, V.; Ghosh, A.; Narayan, C.; Ramya, V. K.; Kaur, P.; Sharma, S.; Werngren, J.; Hoffner, S.; Panduga, V.; Kumar, C. N. N.; Reddy, J.; Kumar, K. N. M.; Ganguly, S.; Bharath, S.; Bheemarao, U.; Mukherjee, K.; Arora, U.; Gaonkar, S.; Coulson, M.; Waterson, D.; Sambandamurthy, V. K.; De Sousa, S. M. Novel N-linked aminopiperidine-based gyrase inhibitors with improved hERG and in vivo efficacy against mycobacterium tuberculosis. J. Med. Chem. 2014, 57, 4889−4905. (25) Wager, T. T.; Pettersen, B. A.; Schmidt, A. W.; Spracklin, D. K.; Mente, S.; Butler, T. W.; Howard, H.; Lettiere, D. J.; Rubitski, D. M.; Wong, D. F.; Nedza, F. M.; Nelson, F. R.; Rollema, H.; Raggon, J. W.; Aubrecht, J.; Freeman, J. K.; Marcek, J. M.; Cianfrogna, J.; Cook, K. W.; James, L. C.; Chatman, L. A.; Iredale, P. A.; Banker, M. J.; Homiski, M. L.; Munzner, J. B.; Chandrasekaran, R. Y. Discovery of two clinical histamine H(3) receptor antagonists: trans-N-ethyl-3fluoro-3-[3-fluoro-4-(pyrrolidinylmethyl)phenyl]cyclobutanecarbox amide (PF-03654746) and trans-3-fluoro-3-[3-fluoro-4-(pyrrolidin-1ylmethyl)phenyl]-N-(2-methylpropyl)cyc lobutanecarboxamide (PF03654764). J. Med. Chem. 2011, 54, 7602−7620. (26) Brodney, M. A.; Beck, E. M.; Butler, C. R.; Barreiro, G.; Johnson, E. F.; Riddell, D.; Parris, K.; Nolan, C. E.; Fan, Y.; Atchison, K.; Gonzales, C.; Robshaw, A. E.; Doran, S. D.; Bundesmann, M. W.; Buzon, L.; Dutra, J.; Henegar, K.; LaChapelle, E.; Hou, X.; Rogers, B. N.; Pandit, J.; Lira, R.; Martinez-Alsina, L.; Mikochik, P.; Murray, J. C.;

Ogilvie, K.; Price, L.; Sakya, S. M.; Yu, A.; Zhang, Y.; O’Neill, B. T. Utilizing structures of CYP2D6 and BACE1 complexes to reduce risk of drug-drug interactions with a novel series of centrally efficacious BACE1 inhibitors. J. Med. Chem. 2015, 58, 3223−3252. (27) Bonomo, S.; Tosco, P.; Giorgis, M.; Lolli, M.; Fruttero, R. The role of fluorine in stabilizing the bioactive conformation of dihydroorotate dehydrogenase inhibitors. J. Mol. Model. 2013, 19, 1099−1107. (28) Tanaka, H.; Shishido, Y. Synthesis of aromatic compounds containing a 1,1-dialkyl-2-trifluoromethyl group, a bioisostere of the tert-alkyl moiety. Bioorg. Med. Chem. Lett. 2007, 17, 6079−6085. (29) Barnes-Seeman, D.; Jain, M.; Bell, L.; Ferreira, S.; Cohen, S.; Chen, X. H.; Amin, J.; Snodgrass, B.; Hatsis, P. Metabolically stable tert-butyl replacement. ACS Med. Chem. Lett. 2013, 4, 514−516. (30) Lobell, M.; Hendrix, M.; Hinzen, B.; Keldenich, J.; Meier, H.; Schmeck, C.; Schohe-Loop, R.; Wunberg, T.; Hillisch, A. In silico ADMET traffic lights as a tool for the prioritization of HTS hits. ChemMedChem 2006, 1, 1229−1236. (31) Hastie, T.; Tibshirani, R.; Friedman, J. The Elements of Statistical Learning; Springer: New York, 2009. (32) Kuhn, M.; Johnson, K. Applied Predictive Modeling; Springer: New York, 2013. (33) A similar molecular weight correction has previously been reported that uses various scaling factors to modify the molecular weight of halogens. See ref 30. (34) Morgenthaler, M.; Schweizer, E.; Hoffmann-Röder, A.; Benini, F.; Martin, R. E.; Jaeschke, G.; Wagner, B.; Fischer, H.; Bendels, S.; Zimmerli, D.; Schneider, J.; Diederich, F.; Kansy, M.; Müller, K. Predicting and tuning physicochemical properties in lead optimization: amine basicities. ChemMedChem 2007, 2, 1100−1115. (35) Smart, B. E. Fluorine substituent effects (on bioactivity). J. Fluorine Chem. 2001, 109, 3−11. (36) Huchet, Q. A.; Kuhn, B.; Wagner, B.; Kratochwil, N. A.; Fischer, H.; Kansy, M.; Zimmerli, D.; Carreira, E. M.; Muller, K. Fluorination patterning: A study of structural motifs that impact physicochemical properties of relevance to drug discovery. J. Med. Chem. 2015, 58, 9041−9060. (37) Shake flask logD. Assay method adapted from published protocol: Hay, T.; Jones, R.; Beaumont, K.; Kemp, M. Modulation of the partition coefficient between octanol and buffer at pH 7.4 and pKa to achieve the optimum balance of blood clearance and volume of distribution for a series of tetrahydropyran histamine type 3 receptor antagonists. Drug Metab. Dispos. 2009, 37, 1864−1870. (38) (a) Bondi, A. van der Waals volumes and radii. J. Phys. Chem. 1964, 68, 441−445. (b) Leo, A.; Hansch, C.; Jow, P. Y. C. Dependence of hydrophobicity of apolar molecules on their molecular volume. J. Med. Chem. 1976, 19, 611−615. (c) Abraham, M. H.; McGowan, J. C. The use of characteristic volumes to measure cavity terms in reversed phase liquid chromatography. Chromatographia 1987, 23, 243−246. (d) Galvez, J. Prediction of molecular volume and surface of alkanes by molecular topology. J. Chem. Inf. Comput. Sci. 2003, 43, 1231−1239. (e) Zhao, Y. H.; Abraham, M. H.; Zissimos, A. M. Determination of McGowan volumes for ions and correlation with van der Waals volumes. J. Chem. Inf. Comput. Sci. 2003, 43, 1848− 1854. (39) Pipeline Pilot, version 9; Accelrys, Inc.: San Diego, CA, 2013. (40) Berman, H. M.; Westbrook, J.; Feng, Z.; Gilliland, G.; Bhat, T. N.; Weissig, H.; Shindyalov, I. N.; Bourne, P. E. The Protein Data Bank. Nucleic Acids Res. 2000, 28, 235−242. (41) Corina, version 3.6; Molecular Networks: Erlangen, Germany, 2014. (42) OMEGA, version 2.5.1.4; OpenEye Scientific Software: Santa Fe, NM, U.S.. (43) Keefer, C. E.; Chang, G.; Kauffman, G. W. Extraction of tacit knowledge from large ADME data sets via pairwise analysis. Bioorg. Med. Chem. 2011, 19, 3739−3749. (44) Assay method adapted from published protocols: Riley, R. J.; McGinnity, D. F.; Austin, R. P. A unified model for predicting human hepatic, metabolic clearance from in vitro intrinsic clearance data in L

DOI: 10.1021/acs.jmedchem.6b00027 J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry

Article

hepatocytes and microsomes. Drug. Metab. Dispos. 2005, 33, 1304− 1311. (45) Obach, R. S. Prediction of human clearance of twenty-nine drugs from hepatic microsomal intrinsic clearance data: An examination of in vitro half-life approach and nonspecific binding to microsomes. Drug. Metab. Dispos. 1999, 27, 1350−1359. (46) Outliers in the data should be treated with caution given the experimental error in logD measurements.

M

DOI: 10.1021/acs.jmedchem.6b00027 J. Med. Chem. XXXX, XXX, XXX−XXX