Molecular Property, Bioavailability, and Permeability Analysis of Bayer's Compound Dataset

Sep 27, 2023 06:30 CST Updated 06:30

Bayer

Pharmaceutical Product R&D Developer

On February 8, 2023, JMC magazine reported on the molecular properties, bioavailability, and permeability data analysis of the Bayer compound series released by researchers from the pharmaceutical development team at the German pharmaceutical giant Bayer. These data and correlation analyses have significant practical implications for drug molecule design. As stated in the article, "These property−structure relationships are combined to provide design guidelines for bioavailable drugs in both traditional and 'beyond rule of 5' (bRo5) chemical space." Based on the detailed and specific content of the original text, the author shares these insights with readers.

The corresponding author of the original text is Daniel H. O’ Donovan, and the other three authors, Claudia De Fusco, Lara Kuhnke, and Lara Kuhnke, are all from Bayer AG in Germany.

The main body of the original text is quite lengthy, fortunately, the author summarized the conclusive key points for optimizing F% (oral bioavailability) and permeability at the end. The tweet will also present these points to everyone at the end.

Compared with other types or routes of administration, oral small-molecule drugs have numerous advantages, which will not be elaborated here. For oral small-molecule drugs, the goal of medicinal chemists is to design compounds with high oral bioavailability (F%) to achieve predictable and robust exposure in clinical settings.

Many factors, such as dissolution from solid to liquid phase (solubility), first-pass metabolism in the liver (hepatic metabolism), intestinal metabolism, and chemical instability, can lead to incomplete bioavailability. However, passage through the intestinal wall (permeability) is the key determinant of F%. These processes can be simulated using in vitro and in vivo tests, enabling drug designers to efficiently address issues related to low F%. In drug chemistry research, problems associated with drug metabolism and pharmacokinetics (DMPK) can be resolved through property-based design. For instance, molecular design teams create new chemical structures to modulate molecular properties such as logD and molecular weight, leveraging established relationships between these properties and DMPK characteristics to guide the optimization of oral bioavailability, permeability, and related parameters.

Lipinski’s Rule of Five (Source: Reference 6)

Several key papers have elucidated the relationship between molecular properties and F%, starting with the groundbreaking "rule-of-five" (Ro5) introduced by Lipinski and colleagues and expanded by Veber et al. In 2001, Lipinski, a senior medicinal chemist at Pfizer, proposed the five fundamental rules for screening drug-like molecules, abbreviated as Rule of 5 (Ro5), also known as the "five rules," since they are all related to the number five. Both authors recommended that molecular properties should be constrained within specified ranges or adjusted to maximize the likelihood of achieving high oral bioavailability. Similarly, authors from Pfizer introduced the "Golden Triangle," outlining the logD and molecular weight ranges to maximize the potential for low clearance and high oral absorption. Other authors have proposed effective computational solutions for predicting DMPK behavior based on chemical structure, molecular properties, or in vitro assay data. Several composite parameters based on molecular properties have been developed to maximize "drug-likeness," among which ligand lipophilic efficiency (LLE) is perhaps the most widely used. LLE combines potency and lipophilicity into a single parameter, while researchers at Pfizer proposed lipophilic metabolic efficiency (LipMetE) as a corollary that integrates hepatic metabolism and lipophilicity. A team from Roche recently demonstrated that LipMetE is proportional to the logarithm of the in vivo half-life (t1/2).

Figure 1. Source: Reference 3

Composite parameters have also been applied to drug design for "beyond Rule of Five" (bRo5) compounds. By using a simple multi-parameter scoring function, DeGoey and colleagues at AbbVie demonstrated that designing orally bioavailable bRo5 compounds is possible. In another paper, colleagues at AstraZeneca attempted to identify chemical properties related to the bioavailability of proteolysis-targeting chimeras (PROTACs), an emerging drug modality with molecular weights and characteristics far exceeding the boundaries of Ro5 space. Authors from Novartis also reassessed the Ro5 concept, noting that as drug properties evolve, prescribed limits become less critical. Similarly, Tinworth and Young from GSK suggested that Ro5 limits, such as molecular weight, may be less important than logD and the number of aromatic rings in predicting permeability and solubility of drug-like compounds. Taken together, these studies indicate that while structure−property correlations and constraints may be useful in molecular design, it is challenging to reconcile Ro5-based design rules with the recent shift in drug discovery within the bRo5 space.

To identify trends in molecular properties that could guide the optimization of oral bioavailability and permeability, Bayer researchers examined data from approximately 2,000 compounds in the Bayer Pharmaceuticals compound collection for which rat bioavailability data had been measured. Using a large amount of available Caco2 permeability assay data (over 20,000 compounds), the permeability data of the Bayer compound set was analyzed and compared with calculated molecular properties such as cLogD7.5, molecular weight, and calculated pKa. Unless otherwise specified, calculated logD values were used throughout the study to maximize the dataset available for analysis. When comparing measured and predicted logD values, the Bayer cLogD7.5 model provided an excellent prediction with a Pearson correlation R² = 0.90.

Figure 2. Experimental and Predicted logD7.5 (Reference 4)

Similarly, for basic and acidic compounds that are partially or fully ionized at physiological pH (approximately 7.5), the advent of reliable pKa calculations has proven to be particularly useful. Researchers at Bayer identified trends between pKa and permeability, which could assist the medicinal chemistry community in optimizing DMPK properties. The researchers incorporated both Ro5 and bRo5 compounds in their analysis to derive overarching trends not limited by subsets of chemical space. Overall, this study summarized established parameters of drug-likeness while also shedding light on new and useful design guidelines based on drug molecular properties to facilitate the optimization of F% and permeability.

Before explaining this article, there are a few basic concepts that need to be reviewed again:

✔Lipophilicity is the tendency of a compound to preferentially partition into nonpolar lipid media rather than into aqueous phases. It plays a crucial role in determining other properties of a drug.

A traditional method for evaluating lipophilicity is to examine the two-phase distribution of a compound between immiscible nonpolar and polar liquids. Typically, octanol is used as the nonpolar phase, while a buffered aqueous solution serves as the polar phase. The measured distribution values are expressed as Log P and Log D.

Log P: The logarithm of the partition coefficient of a compound, under pH conditions where all molecules exist in neutral form, between an organic phase (e.g., octanol) and an aqueous phase (e.g., buffer) is referred to as Log P.Log P = Log([Compound Organic Phase]/[Compound Aqueous Phase]) Log D: At a specific pH (x), the compound exists partially in ionized form and partially in neutral molecular form; the logarithm of the distribution coefficient of the compound at this pH between an organic phase (e.g., octanol) and an aqueous phase (e.g., buffer) is Log D. Log DpHx = Log([Compound Organic Phase]/[Compound Aqueous Phase]).

The Log P value depends on the distribution of neutral molecules in the two phases, while the Log D value depends on the distribution of molecules in their neutral form plus the distribution of molecules in their ionized form. Ions have a greater affinity for the polar aqueous phase than for the non-polar organic phase. The proportion of ionized molecules depends on the pH of the aqueous solution, the pKa of the compound, and whether the compound is an acid or a base. For acidic compounds, the ratio of neutral molecules to anions in the solution decreases as the pH increases, so the Log D also decreases as the pH increases. Conversely, for basic compounds, the ratio of neutral molecules to cations in the solution increases as the pH increases. Therefore, the Log D also increases as the pH increases.

✔Drug molecules encounter various membrane barriers in living systems, including gastrointestinal (GI) epithelial cells, capillary walls, hepatocyte membranes, glomeruli, restrictive organ barriers [such as the blood-brain barrier (BBB)], and target cell membranes.

The permeability of a compound varies depending on the membrane, and this difference arises from variations in the lipid system of the cell membrane (passive diffusion), the expression of membrane transport proteins (active transport), or the tightness of cell junctions (paracellular pathway).

The membrane permeability of compounds involves different mechanisms, including passive diffusion, active transport, endocytosis, efflux, and paracellular pathways.

Main Membrane Permeability Mechanisms (Source: Reference 6)

The permeability of a compound is the integration of all feasible permeation mechanisms mentioned above. The term "absorptive transport" is often used to describe the permeation of a compound from the gastrointestinal tract into the bloodstream. As shown in Figure 2, absorptive transport results from the combined effects of passive diffusion (driven by concentration gradients and pH effects), active transport (driven by transporter affinity), and paracellular permeation (driven by molecular size, polarity, and concentration gradients). Conversely, the term "secretory transport" is commonly used to describe the permeation of a compound toward the gastrointestinal tract, which results from the combined effects of passive diffusion and efflux.

✔Commonly used Caco-2 cells (human colon cancer epithelial cells) to measure drug permeability, conduct drug permeability experiments, and study the bidirectional transport of drugs. By examining factors such as time and drug concentration, the apparent permeability coefficient of the drug is calculated. Caco-2 cells are very similar in morphology to small intestinal epithelial cells, possessing the same cell polarity and tight junctions. Various transport systems and metabolic enzymes found in small intestinal epithelial cells are also present in Caco-2 cells.

There are 2,237 compounds with rat oral bioavailability data available, representing a variety of chemical structures obtained or synthesized by Bayer AG during its pharmaceutical research and development activities. Although medicinal chemists tend to prioritize more "drug-like" and in vitro-optimized compounds for in vivo pharmacokinetic studies, leading to some bias in structural and property diversity, the Bayer dataset still encompasses various structural classes from different projects.

The research focused on three key parameters (solubility, liver metabolism, and permeability) to determine their impact and correlation with oral bioavailability. Researchers from Bayer Pharmaceuticals first addressed solubility because it plays a critical role in bioavailability—compounds must dissolve to be absorbed. However, using kinetic solubility data (PBS or citrate buffer medium, pH 6.5), zero correlation was observed with rat bioavailability (R² < 0.01, n = 2018 compounds). The researchers were not surprised by this result since kinetic solubility is merely an approximation of the more relevant thermodynamic solubility parameter. Although kinetic solubility can help exclude in vitro assays potentially affected by compound precipitation, its correlation with oral bioavailability may be low. Moreover, in vivo pharmacokinetic studies are typically conducted using pre-dissolved amorphous substances at low doses, which reduces the influence of dissolution rate and solubility on measured F%. Rapid pH changes during intestinal transit also aid in compound dissolution, especially for compounds that may be ionized at lower pH levels. Based on these considerations, drug discovery teams often measure solubility in simulated gastric and intestinal fluids or at different pH levels to better assess the impact of solubility under biologically relevant conditions.

Study on the Correlation Between Caco2 Cell Permeability (Apparent Permeability Coefficient, Apical to Basolateral, i.e., Papp A-B), Rat Hepatocyte (Rat Heps) Metabolic Stability, and F% is More Valuable. The Linear Regression of F% with Papp A-B and Rat Heps on a Logarithmic Scale Shows Moderate Correlation (R² = 0.26). Heatmap Plots of These Three Parameters Reveal a Clear Trend Consistent with the Expected Impact of Permeability and Liver Metabolism on F%. Empirically, Intestinal Permeability and Drug Efflux are the Most Successful DMPK Optimization Parameters in Preclinical Candidate Selection. Among All Preclinical Candidate Compounds Used for Oral Administration at Bayer, Not a Single Example of Poor Permeability in the Caco2 A-B Papp Assay Has Been Found.

Figure 3. The relative importance of eight commonly used molecular property descriptors as variables in a linear regression model with respect to the logarithm of metabolic stability in rat hepatocytes (L/h/kg). (Reference 1)

Although hepatic metabolism is a key determinant of bioavailability, the structure−property relationships behind metabolism have been extensively covered in previous reports, i.e., among commonly used molecular properties, lipophilicity may be the strongest predictor of hepatic metabolism, as it is thought that lipophilic compounds bind more strongly to metabolic enzymes through hydrophobic interactions. Consistent with this effect, a regression analysis examining eight commonly used molecular properties identified cLogD7.5 as the most important single variable determining rat hepatocyte stability. This analysis was performed using a larger dataset (n = 21,163 compounds) compared to oral bioavailability data. The number of rotatable bonds (rotBonds), fraction sp3 (F(sp3)), and topological polar surface area (TPSA) were the next most significant variables in the researchers' analysis. This result is consistent with previous authors describing the importance of these properties in determining hepatic clearance.

It should be noted that the predictive power of this regression analysis is limited, and several of these attributes may also be interrelated; for example, compounds with higher molecular weight tend to have higher lipophilicity. The relationship between these properties and liver metabolism is also not strictly linear. Nevertheless, this simple model highlights the importance of lipophilicity in determining hepatic clearance, with other molecular properties contributing less significantly.

Next, the researchers attempted to analyze the relationship between calculated properties and the permeability measured in Caco2 cell assays. Throughout the study, compounds with recovery rates below 65% in the Caco2 assay were excluded to ensure only meaningful data was included. Low solubility may be a factor contributing to low recovery rates, as 49% of the excluded compounds showed solubility ≤1 mg/L. In contrast, only 23% of the included compounds (Caco2 recovery rate above 65%) displayed the same level of low solubility.

Regression Analysis Review of Caco2 Papp A-B with the Eight Attribute Descriptive Parameters in Figure 3 Above: Lipophilicity is once again emphasized as the most important determinant of permeability, closely followed by molecular weight, and then hydrogen bond donor (HBD) count. Given the importance of lipophilicity to permeability, the figure below explores the relationship between cLogD7.5 and Papp A-B. To examine the interaction with molecular weight, researchers categorized compounds into low (≤400), medium (400 < x ≤ 600), and high (x > 600) molecular weight ranges. The resulting graph highlights the expected trend that smaller molecules benefit from higher average permeability. In extreme cases of low (cLogD7.5 < 0.5) and high lipophilicity (cLogD7.5 > 5), the average permeability approaches zero. However, outliers do exist, particularly among high logD compounds (cLogD7.5 > 5.5), where several permeable examples can be found.

Figure 4. (Left) Relationship between cLogD7.5 and the geometric mean of Caco2 Papp A-B (nm/s) (Reference 1)

Notably, the researchers observed a "sweet spot" for compounds with a cLogD7.5 of 3.25, which exhibited the highest average permeability. This value remained constant, regardless of the molecular weight range considered. The result is highly consistent with the optimal cLogD=3 value proposed by DeGoey et al. in their analysis of high molecular weight bRo5 compounds. Meanwhile, an analysis by Bayer researchers indicated that this optimum value remains unchanged even for smaller compounds with a molecular weight ≤ 400.

Researchers also investigated the relationship between lipophilicity and efflux ratio (ER) in Caco2 permeability assays. Consistent with previous observations on Papp, a trend of lower efflux ratios for compounds with lower molecular weight was noted. However, contrary to the relationship with Papp, there were differences in cLogD7.5 values when the efflux ratio approached its maximum, in which case the efflux ratio varied with molecular weight.

Figure 5. Relationship between molecular weight and maximum efflux ratio in Caco2 permeability assay (Reference 1)

The fact that the cLogD7.5 value with the highest average efflux varies with molecular weight contrasts with the previously observed situation for Papp, which consistently reaches its maximum permeability at cLogD7.5=3.25. Based on this result, Bayer researchers recommend that drug design teams differentially optimize the logD value depending on whether the issue is related to Papp A-B or efflux.

Bayer researchers further analyzed the impact of molecular weight and logD on A-B permeability through heat maps. Pfizer's Golden Triangle and Waring from AstraZeneca both proposed boundaries for these two parameters to maximize the likelihood of oral absorption and high permeability. Young and colleagues from GSK established a related approach, suggesting that calculating molar refraction (a characterization of molecular size) and logD7.4 is beneficial for optimizing permeability.

The analysis by Bayer researchers largely summarized the same trend identified by previous authors, indicating that compounds with extreme logD values require lower molecular weight to achieve high permeability. In contrast, compounds with cLogD7.5 values close to the "sweet spot" of 3.25 can maintain good permeability within a molecular weight of 700. This optimal logD value differs from the Golden Triangle, where the logD showing maximum tolerance for higher molecular weight compounds is significantly lower (between 1.0 and 2.0). A careful review of Pfizer's authors' original publication, in which they constructed the Golden Triangle while analyzing permeability, reveals that the apparent optimal logD for the permeability they noted falls between 3.0 and 4.0.

Golden Triangle studied the impact of logD and molecular weight on oral absorption, while Waring's analysis focused on permeability. When examining data collected by AstraZeneca, Waring proposed the minimum logD required for high permeability (defined as Papp > 100 nm/s) within different molecular weight ranges. Using Bayer's pharmaceutical dataset, it was found that these guidelines could successfully predict high-permeability and low-permeability compounds with accuracies of 63% and 70%, respectively. Waring’s guidelines suggest that larger molecules require higher lipophilicity to achieve permeability. However, this linear relationship cannot be maintained at high lipophilicity (logD > 3.5–5.0), and beyond this range, permeability begins to decline as logD increases. The authors of Golden Triangle had previously noted the same trend, hypothesizing that this effect might be due to membrane solvation, where high logD compounds are retained in the lipid membrane, thereby hindering permeability.

Figure 6. Waring Guiding Principles for the Analysis of the Ability to Differentiate Hypertonic Compounds from Hypotonic Compounds (Reference 4)

Bayer pharmaceutical researchers further analyzed the relationship between calculated pKa, molecular weight, and permeability measured in the Caco2 assay.

✔Here, let me briefly introducepKa，Acid coefficient, also known asAcid Dissociation Constant, Code NameKa value, in chemistry andBiochemistryIn China, it refers to a specific equilibrium constant, denoted as Ka, representing an acid.DissociationThe ability of hydrogen ions.When the concentration of drug ions and non-ions in the solution are completely equal, each accounting for 50%, the pH value of the solution is called the dissociation constant of the drug.The relationship between pKa and pH is: pKa = pH + log([Undissociated Concentration]/[Dissociated Concentration]).

Several authors also commented on the relationship between protonation state and permeability. However, there are currently no large-scale studies aimed at identifying general trends in this property. This may be due to the limited data obtained from low-throughput pKa measurement experiments. Fortunately, the emergence of accurate pKa prediction software has made calculated pKa a parameter for constructing structure-property relationships. To this end, Bayer researchers adopted a commercial prediction model jointly developed by Bayer and Simulations Plus.

Figure 7 shows a heatmap that investigates the trend between the predicted pKa of the most basic center, molecular weight, and the average permeability of 24,105 compounds measured by the Caco2 Papp A-B assay, encompassing 10-90% of Caco2 permeability. The upper and lower percentile ranges provide an estimate of the permeability limits that can be expected for a given combination of molecular weight and pKa. The dataset may be biased toward including more "drug-like" compounds since these are more likely to be selected for Caco2 assays. Nonetheless, these compounds represent a variety of molecular structure choices and should help infer general trends.

Figure 7. (Left) Heatmap plot of Caco2 Papp A-B (nm/s) versus molecular weight (Da) and the predicted pKa of the most basic center for 24,105compounds. (Source: Reference 1)

Figure 7 confirms the trend that Papp A-B permeability becomes more challenging as molecular weight increases. Investigating the interplay between molecular weight and pKa, the dataset can be subdivided into non-basic compounds, which are unprotonated at physiological pH 7.5 (i.e., the upper half of the heatmap, compounds with pKa ≤ 6), and basic compounds, which are expected to be partially or fully ionized at physiological pH (pKa > 6). This subdivision into basic and non-basic compounds reflects the profound impact of charge state on passive diffusion permeability, with neutral/uncharged molecules benefiting from higher affinity and diffusion into the lipid bilayer of cell membranes.

Study of the average Papp A-B values reveals that basic compounds (pKa > 6) exhibit good permeability at molecular weights up to around 500, beyond which their average permeability becomes significantly poorer. This threshold is reminiscent of Ro5's recommendation for maximum molecular weight. However, examining the 90th percentile plot shows that even at higher molecular weights of up to 700, there may still be permeable basic compounds. Above this value, there are far fewer basic compounds with good permeability. Compounds with stronger basic centers (pKa 9-10) also show a clear trend of displaying good average permeability only at lower molecular weights (<400), likely due to their cationic protonated state hindering diffusion through the lipophilic membrane. Although increased lipophilicity might enhance the permeability of these highly basic molecules, such a design strategy could raise the risk of drug-induced phospholipidosis. In addition to passive diffusion, protonated molecules can also cross the phospholipid bilayer via mechanisms specific to such ions, including degradation of the lipid structure of the membrane. Unlike passive diffusion, this degradation transport mechanism appears to be structure-dependent rather than driven by lipophilicity. Overall, achieving high permeability in high molecular weight strong bases is expected to be highly challenging.

Fig. 8. Heatmap plot of Caco2 apparent permeability A-B (nm/s) versus molecular weight (Da) and the predicted pKa of the most acidic center for 22,559 compounds. (Source: Reference 1)

Using the data shown in Figures 7 and 8, the optimal molecular weight and pKa ranges for A-B permeability can be determined. These ranges depend on the ion category and can provide more detailed guidance by subdividing the data into strong acids, weak acids, weak bases, neutral compounds, and zwitterions. Zwitterions generally exhibit poor passive permeability, but highly permeable zwitterionic compounds are commonly found by limiting the pKa range of their ionization centers, even up to a molecular weight limit of 600. Researchers further observed that regardless of molecular weight, zwitterions containing strongly acidic or strongly basic centers typically exhibit poor A-B permeability.

✔Here, we introduce the abbreviations of several terms: %ABS: Absorption Percentage; TPSA: Topological Polar Surface Area; NROTB: Number of Rotatable Bonds; HBA: Number of Hydrogen Bond Acceptors; HBD: Number of Hydrogen Bond Donors.

Molecular weight is not the only reason for poor permeability. As compounds become larger, design teams introduce more molecular flexibility and polarity characteristics (hydrogen bond donors, logD-lowering functional groups, etc.) to improve drug-likeness and capture additional interactions with target molecules. These features not only hinder permeability through passive diffusion but may also lead to additional recognition elements for transporters, enhancing their efflux capability. Therefore, molecular weight can be considered a representative of multiple potential characteristics affecting permeability. To better understand the role of these factors, researchers at Bayer investigated the impact of further molecular properties on permeability. Figure 9 illustrates the relationships between HBD count, HBA count, TPSA, rotatable bonds, aromatic rings, fraction sp3, and Caco2 Papp A-B permeability and efflux. In almost every case, as the molecular characteristic values increase, the average Papp A-B permeability decreases significantly. This reflects the same effect observed with molecular weight — as compounds become larger and include more structural features, their permeability tends to decrease.

Fig. 9. Line plots of molecular properties (HBD, HBA, TPSA, rotatable bond count, aromatic ring count, and fraction sp3) versus Caco2 Papp A-B (nm/s), showing the geometric mean and the 10th to 90th percentile range in orange and the geometric mean of the Caco2 efflux ratio in blue. Values for TPSA are binned per interval of 50 Å², and values for F(sp³) are binned per interval of 0.05. (Source: Reference 1)

Regarding Figure 9 above, the editorial team of "Pharmacological Discussion" briefly highlights a few points: For HBA counts, the A-B permeability decreases gradually between values of 1 and 10, and above this range, Papp approaches its minimum value. These ranges summarize Lipinski’s Ro5 guidelines, while compounds with HBA = 0 are primarily quaternary ammonium salts, a class of compounds with poor permeability. Compared to several other properties, the relationship between F(sp3) and Papp A-B or efflux rate is much weaker. Researchers from Bayer also plotted a linear graph of basolateral-to-apical cell membrane permeability, which follows the same general trend shown in Figure 9.

Figure 10.B-A Permeability versus Molecular Properties (Source: Reference 4)

The strength and importance of the effects of these parameters on Caco2 cell permeability can be explained by their synergistic effects on passive permeation and efflux, which together enhance the impact on net permeability. The relationship between these properties and the Caco2 efflux ratio follows an inverse trend, meaning that as molecular property values increase, efflux increases. At higher values of HBD, HBA, TPSA, rotBonds, and the number of aromatic rings, the efflux ratio appears to decrease after surpassing a certain maximum value. For example, in the case of HBD statistics, the average efflux ratio reaches its maximum at HBD=5, above which the efflux ratio begins to decline. However, this may be due to Papp B-A and Papp A-B approaching very low values at high HBD values, resulting in an apparent efflux ratio of approximately 1. For similar reasons, the unstable behavior observed in the efflux ratio plots for higher HBA values and rotatable bond counts may also be an artificial effect when Papp A-B and B-A values approach the lower detection limit. Additionally, in extreme cases, fewer compounds are available for analysis, leading to greater data noise.

To some extent, the analysis by Bayer researchers generalizes the constraints of Ro5 chemical space. However, there are significant exceptions to the 500 molecular weight limit in reality, especially for uncharged compounds, which maintain good average permeability at molecular weights up to 700. This is consistent with the "extended rule of five" (eRo5), defined by Kihlberg and colleagues as compounds with a molecular weight of 500-700, HBD ≤ 5, HBA ≤ 10, PSA ≤ 200, and the number of rotatable bonds ≤ 20. Within the Ro5 and eRo5 chemical spaces, the study by Bayer researchers confirms that the aforementioned trends are useful predictors of permeability.

The analysis of bRo5 compounds is affected by the limitations of molecular weight in existing datasets, where most compounds have a molecular weight below 700, and few exceed this range. In the studied dataset, only 586 compounds with a molecular weight ≥700 were tested in Caco2 experiments, making it difficult to infer statistically significant trends within this range. Among these high-molecular-weight compounds, over 90% exhibited low Papp A-B permeability. This contrasts sharply with compounds having a molecular weight below 500 and eRo5 in the 500–700 range, where less than 50% showed poor permeability.

Figure 11. Bar chart showing the distribution of Caco2 apparent permeability A-B (nm/s) values for compounds with molecular weights in the Ro5-like (<500), eRo5-like (500-700), and bRo5-like (≥700) categories. (Reference 1)

As shown in the figure above, compounds with a molecular weight >700 generally exhibit poor permeability. Of the 586 such compounds (>700), only 36 showed moderate or good Papp A-B permeability (≥30 nm/s). Among these 36 compounds, shared characteristics were identified: 10 of them can be described as macrocyclic compounds containing macrocyclic structures. Another feature present in 8 of the 36 compounds is a structural element that can be classified as a linker connecting two chemical moieties. Linker design is a rapidly evolving field with specific applications in PROTAC development. The incorporation of a carefully designed linker allows bRo5 molecules to adopt conformations that mask their polarity in a manner analogous to macrocyclization, and it also reduces the radius of gyration. A cross-sectional area threshold of 80 Å² is critical for tilting the balance toward passive diffusion; this threshold is high enough to effectively shorten the membrane residence time of drug molecules, whereas below this threshold, efflux pumps can easily extract molecules from the membrane. This characteristic has recently been termed "molecular chameleonicity" and may provide higher-than-expected permeability for PROTACs and other bRo5 compounds. Drug design teams seeking to incorporate chameleonic properties could benefit by studying the behavior of their compounds in aqueous and non-polar solvent environments; a parameter called "chameleonic efficiency" has been proposed for this purpose. Another useful tool for analyzing bRo5 molecular permeability is the experimental polar surface area (EPSA) method, a chromatographic technique developed by Pfizer for analyzing intramolecular hydrogen bonding. EPSA technology has been successfully applied to optimize compounds with permeability challenges, including peptides and other large polar molecules.

Given the challenges of achieving good DMPK properties for bRo5 compounds, Bayer researchers actively evaluated design guidelines reported in the literature. In a recent article, DeGoey and his colleagues at AbbVie proposed the AB-MPS score as a comprehensive parameter to guide the optimization of bioavailability for bRo5 compounds. The authors defined bRo5 compounds as those violating more than one Lipinski rule and having a molecular weight exceeding 500. Using the formula AB-MPS = (abs(cLogD7.4 − 3) + aromatic ring count + rotatable bond count), bRo5 compounds with an AB-MPS score >14 showed a tendency toward poor oral bioavailability in rats. Figure 12 displays the AB-MPS scores applicable to Bayer compounds meeting AbbVie’s bRo5 criteria, categorized into quartile ranges of F% according to the original publication. Compared to the AbbVie dataset, the researchers’ selection included fewer compounds (n = 375) and a narrower range of AB-MPS scores, from 23.1 to 6.3, with a notable difference between the used cLogD7.5 and measured logD7.4. Nevertheless, Bayer researchers were keen to evaluate this scoring function using data from their own company.

Figure 12. AB-MPS box plots for compounds with >1 Lipinski Ro5 violations and molecular weight >500, showing oral F% in the rat versus AB-MPS value, with compounds binned by quartile distribution of their bioavailability values. (Source: Reference 1)

The researchers confirmed the overall trend that compounds with poorer bioavailability have higher AB-MPS values, as indicated by the average scores in the figure above. Using a specified cutoff of AB-MPS >14 and a bioavailability threshold of F% >20, the scoring function was able to correctly classify 65% and 52% of the compounds as having good/poor bioavailability, respectively. Further comparison of the AB-MPS scores was made with calculated FaFg values, which measure intestinal absorption in the absence of hepatic metabolism interference, and the results were similar to F%.

Figure 13. Percentage of compounds correctly classified as achieving acceptable bioavailability (F% > 20) using different bRo5 design guidelines (Reference 1)

Bayer researchers also considered two property-based design rules proposed by Veber et al., both of which can be used to help achieve oral bioavailability. What is notable about these rules is that they exclude molecular weight criteria, allowing them to be applied to both Ro5 and bRo5 chemical spaces. The first Veber rule suggests a critical value of PSA≤140Å² and a rotatable bond count ≤10. When applied to the entire dataset of Bayer pharmaceutical research (Ro5 + bRo5 compounds), these thresholds correctly predicted 67% and 53% of compounds with good/poor bioavailability, respectively. The second Veber rule limits the sum of HBD+HBA to ≤12, again combined with a rotatable bond count ≤10. These criteria performed slightly better, correctly predicting 67% and 62% of compounds with good/poor bioavailability. These rules also performed well when applied to the subset of bRo5 compounds defined using AbbVie standards. Empirically, all these methods appear to predict F% with reasonable accuracy, suggesting that bRo5 compounds can still be optimized by adjusting their molecular properties.

Finally, the original text summarizes the guidelines or principles for optimizing F% and permeability:

✔Oral bioavailability depends on permeability and metabolic stability. The impact of adjusting any of these parameters on F% can be roughly estimated by referring to the figure below.

Figure 14. (Left) Heatmap of oral bioavailability (F%) in rats for 1930 compounds versus in vitro Caco2 apparent permeability (Papp) and in vitro metabolic stability in rats (A-B). (Right) The same plot colored according to the number of compounds within each data range. (Source: Reference 1)

✔When faced with poor metabolic stability, the design team may address the issue by reducing lipophilicity, as logD is by far the most influential property on metabolic stability. This approach can be supplemented by metabolite ID studies and the identification of metabolic soft spots through structural modifications.

✔Based on experience, the utility of dynamic solubility measurements in guiding bioavailability optimization is limited. Evaluating the impact of solubility on F% requires thermodynamic solubility data, which is often unavailable in the early stages of drug discovery.

✔For compounds with poor permeability, the properties that have the greatest impact are logD, molecular weight, and HBD count. Therefore, permeability may improve when logD is close to 3.25, and both molecular weight and HBD count should be minimized.

✔In addition to the above three parameters, reducing the values of HBA, TPSA, rotatable bonds, and aromatic rings is also a feasible strategy to improve permeability and efflux. In contrast, F(sp3) has less impact.

✔Compounds containing acidic or basic centers can benefit from moving towards a more neutral pKa value to enhance passive permeability. Figures 7 and 8 can be referenced to estimate the potential impact. However, this strategy is not recommended when high efflux results in poor permeability.

✔For larger molecules (molecular weight of 500-700), the pKa of basic compounds should be below 9, and the pKa of acids should not exceed 4 to achieve good passive permeability. Compounds with pKa values exceeding this range can be limited to a lower molecular weight (400-500) to achieve passive permeability. For zwitterions, if the pKa of the basic center is less than 9 and the pKa of the acidic center is greater than 4, a high Papp is possible.

✔These guidelines are most applicable to compounds with a molecular weight <700. Beyond this range, achieving passive permeability in Caco2 assays is often challenging. In such cases, design strategies like macrocyclization or forming intramolecular hydrogen bonds may be helpful. It should also be noted that Caco2 analysis (and other discovery assays) may not provide reliable data for such larger compounds. Other experiments, such as EPSA, might offer more useful information.

Reference:

1.https://doi.org/10.1021/acs.jmedchem.2c01577.

2.Egbert, M.; Whitty, A.; Keseru, G. M.; Vajda, S., Why Some Targets Benefit from beyond Rule of Five Drugs. Journal of medicinal chemistry 2019.

3.Cecere, G.; Guasch, L.; Olivares-Morales, A. M.; Umehara, K.; Stepan, A. F. LipMetE (Lipophilic Metabolism Efficiency) as a Simple Guide for Half-Life and Dosing Regimen Prediction of Oral Drugs.ACS Med. Chem. Lett. 2022, 13, 1444−1451.

4.https://pubs.acs.org/doi/suppl/10.1021/acs.jmedchem.2c01577/suppl_file/jm2c01577_si_001.pdf

5.https://www.chemico.com.cn/h-nd-33.html

6.Halford, B. Wrestling with Lipinski’s rule of 5. Chemical & Engineering News. 2023, 101.

7.Lipinski, C. A. Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Advanced Drug Delivery Reviews. 1997, 23, 3-25.

Disclaimer: The publication/reposting of this article is solely for the purpose of information dissemination, and does not represent the views of this official account or confirm the authenticity of its content. Any judgment made based on this content will be at your own risk.If there is any infringement, please inform us and we will delete it immediately!

Long press to follow this official account

Fan Group/Submission/Authorization/Advertisementetc.

Please contact the official account assistant.

If you think this article is good-looking, please click here ↓