The Emerging Role of A.D.M.E. in Optimizing
Drug Discovery and Design

Robert J. Guttendorf, Ph.D.

Pharmacokinetics/Drug Metabolism Department
Parke-Davis Pharmaceutical Research
Division of Warner Lambert Co.
2800 Plymouth Rd.
Ann Arbor, MI 48105


Though traditionally relegated to providing data and information in support of regulatory filings for new drugs, scientists in pharmacokinetics, drug metabolism, drug absorption, and related disciplines have seen their breadth of responsibility burgeon beyond the realm of drug development. Impelled by exciting technologic advances, an enhanced efficiency in gathering absorption, distribution, metabolism, and excretion (ADME) data has been realized, permitting ADME scientists to contribute more effectively to the drug discovery process as well.

ADME in Drug Development

The term ADME is typically used in reference to nonclinical studies. In reality, ADME is generic, being applicable without restriction to pharmacokinetic/metabolism investigations in humans as well as animals. Fundamentally, ADME information is critical in all phases of a fully integrated drug development program. The initial charge of early development programs is to file the Investigational New Drug (IND) application such that approval may be secured to investigate a new chemical entity (NCE) in humans.

Animal toxicology studies comprise the foundation of the IND. The toxicology program is designed to identify toxicities that may potentially be encountered in humans, helping ensure that the initial human studies are conducted safely and ethically. In this context, ADME studies provide supportive information to augment the interpretation of toxicology findings. Of primary importance among them are toxicokinetic studies, in which systemic drug and/or metabolite exposure in toxicology animals is evaluated. Drug exposure, expressed in terms of AUC (area under the drug plasma concentration-time curve), Cmax (maximum drug concentration in plasma), or an alternative parameter, is then related to dose level and toxicological outcomes [1]. Based on toxicokinetic data at the no-observed toxic effect dose, an acceptable exposure limit in humans can be defined.

To assist in putting toxicokinetic data into a broader perspective, the basic pharmacokinetic behavior of the NCE is assessed in the toxicology species. Some typical studies that may be conducted in this stage of development are listed in Table 1.

Pharmacokinetics/absolute bioavailability in toxicology species (male/female)
Protein binding
Erythrocyte/plasma distribution
Whole body autoradiography/tissue distribution
Mass balance in toxicology species
Metabolite profile in toxicology species
Allometric scaling

In later development, the focus of ADME work is shifted to human studies. These are intended to more fully define the disposition of the drug in humans, particularly in the target therapeutic population. Ultimately these data are integrated into a New Drug Application (NDA) or Market Authorization Application (MAA) to secure final approval to market the drug. Possible studies which may be performed during clinical development are listed in Table 2.

Toxicokinetics in chronic and reproductive/teratology studies
Multiple dose pharmacokinetics in toxicology species
Biliary excretion/enterohepatic recirculation
Metabolite identification in toxicology species
Multiple dose whole-body autoradiography/tissue distribution
Placental transfer
Milk secretion
Effects on metabolic systems (induction)
Single/multiple dose pharmacokinetics in safety/tolerance studies
Dose proportionality
Food effect
Repeated measures (intrasubject variability)
Mass balance/metabolite profile and identification with radiotracer
Pharmacokinetics in subpopulations (gender, age, genetics, liver and renal failure)
Drug interactions
Milk penetration
Population pharmacokinetics/pharmacodynamics in patients
Pivotal bioequivalence

ADME data gathered during the full course of development are incorporated into drug labeling, which is intended to optimize therapeutic utilization of the drug in the target population.

ADME in the Drug Discovery Process

While scientists in the ADME disciplines have historically provided support to drug discovery teams, the unique expertise which they could bring to the discovery process has heretofore been vastly underutilized. In no small way this may be ascribed to the less than expedient experimental procedures required to conduct most ADME studies. The time- and labor-intensiveness of much of this work is not an issue during drug development, where resources can be focused on one compound. This modus operandi does not conform, however, to current drug discovery efforts which have exponentially enhanced their operating efficiency with mass-screening techniques. However, the advent of a host of exciting new technologies is providing hope that scientists in various of the ADME disciplines may now be able to participate more actively and productively in the drug discovery process.


Insofar as the overwhelming majority of new drugs are intended to be administered orally, with few exceptions the ability of an NCE to provide activity by the oral route (i.e. have good bioavailability) is imperative. Whereas oral bioavailability can be determined relatively easily in animal pharmacokinetic studies, it is not reasonable to expect that this approach can function optimally in a large scale screening environment. Fortunately a number of in vitro and cell culture techniques have evolved in recent years that have facilitated the assessment of intestinal permeability.

It has been demonstrated that membrane permeability can be predicted for some compounds with reasonable accuracy based solely on physicochemical parameters. Therefore, close scrutiny of the chemical structure may provide valuable basic information about an NCE prior to the commencement of laboratory experiments. It is well established, for instance, that efficient oral absorption will occur only after drug has dissolved and presented itself to the intestinal mucosal surface from whence it can traverse the epithelium. Dissolution is determined by the highly interdependent influences of aqueous solubility, ionizability (pKa), and lipophilicity (octanol/water log P or log D7.4). Furthermore, log P is a crucial factor governing passive membrane partitioning, influencing permeability opposite to its effect on solubility (i.e. increasing log P enhances permeability while reducing solubility). In light of this counterdependence, it has been suggested that oral absorption may be optimal within a log P range of 0.5 to 2.0 [2].

While the promise of predicting membrane permeability from chemical structure alone is enticing, these methods do not yet enjoy the level of sophistication required to supplant experimental methods. In that regard, a number of in vitro tools have been adapted, with notable efficiency, to high-throughput assessments of membrane permeability and potential oral bioavailability. Most notable among them are CaCO-2 cells. Derived from a human colon carcinoma cell line, these cells are grown in a confluent monolayer on porous membrane filters which are mounted in diffusion chambers. Permeability measurements are based on the rate of appearance of test compound in the receiver compartment. The apical (donor) surface of the monolayer contains microvilli and thus retains many characteristics of the intestinal brush border. The cells also express functional transport proteins [3,4] and metabolic enzymes [5], the degree of expression being dependent upon the post-seeding age of the cells.

Everted intestinal rings and brush-border membrane vesicles (BBMV) are also commonly used systems for assessing membrane permeability. The former technique is a refinement of one of the earliest in vitro absorption systems in which an everted intestinal segment was suspended in a buffer system to measure mucosal-to-serosal transfer. The current methodology involves isolating a rat intestinal segment, everting, and slicing into rings which are suspended in buffer. BBMV are prepared by removing the brush-border surface from rat or rabbit intestine and molding it into vesicles by homogenization and differential centrifugation. Both everted intestinal rings and BBMV are most useful for determining drug uptake rates rather than transepithelial flux.

In the in situ intestinal perfusion system, an intestinal segment is exposed in an anesthetized rat and drug solution is perfused through the lumen in a single-pass or recirculating fashion. Drug permeability is derived from the rate of disappearance of drug from the perfusate. Though more labor-intensive, in situ intestinal perfusions remain popular owing to the perceived clinical relevance of permeability data derived therefrom. In a recent study, it was demonstrated with a series of small organic molecules as well as a series of peptidomimetics that CaCO-2 cells, everted intestinal rings, and in situ perfusions have strong potential for predicting fraction absorbed in humans [6].

One of the most appealing attributes which these experimental systems possess is their capability to perform relatively high throughput screening. This is particularly relevant for CaCO-2 cells, everted intestinal rings, and BBMV. An imposing rate-limiting step, once the systems are established and optimized, is the development of assay methods (usually HPLC) to quantify the analytes of interest. On the positive side, as these experiments are conducted in aqueous buffers, the preanalytical sample purification requirement is minimal. These analytical burdens can be even further reduced by immobilized artificial membranes (IAM), a recently developed system which has been greeted with considerable enthusiasm [7]. Briefly, the analyte is injected onto a specialized IAM HPLC column packed with a phosphatidylcholine (PC) stationary phase selected to closely mimic biological membranes. Theoretically, the chromatographic capacity factor for the analyte should correlate with the water-to-PC membrane partition coefficient, a useful parameter for calculating membrane permeability. The ability of IAM technology to predict membrane permeability is currently being evaluated. Elegant in its simplicity, it is well suited to a rapid screening environment, although it is recognized that its applicability will be limited to those compounds which are absorbed by passive processes.


The systems described above have the greatest utility when absorption is rate-limiting to systemic availability. For many compounds, even if absorption is optimized bioavailability may be limited by extensive metabolism. Indeed, metabolism can complicate the in vivo activity profile irrespective of route of administration.

As with absorption, valuable metabolic input can be imparted to the drug discovery team based on information gleaned from chemical structure alone. It is felt that the teleological underpinnings of xenobiotic transformation reside in modifying environmental toxins so as to facilitate their removal from the body. Not surprisingly, the majority of drug biotransformation processes appear to follow this paradigm, typically resulting in increasingly water soluble metabolic products which can be more readily excreted in urine. Thus one may surmise that within a group of compounds, highly lipophilic drugs would be metabolized most actively. More explicitly, the metabolic potential of a compound containing a chemical moiety known to be avidly biotransformed may be predicted reasonably accurately on the basis of abundant historical data. For example, being cognizant that unhindered phenolic hydroxyl groups are exquisitely good substrates for conjugating enzymes, NCEs which contain this moiety can be eliminated from further consideration if a predilection to rapid metabolism is felt to be detrimental. As we gain information about the specific conformational constraints imposed by the catalytic sites of major metabolic enzymes, our ability to predict in vivo metabolic events based solely on chemical structure is enhanced commensurately.

Given the current state of knowledge, however, it is likely that drug discovery teams will continue to prefer metabolism-related input predicated on experimental data rather than theoretical predictions. To this end, various in vitro methods are available which are being increasingly incorporated into drug discovery strategies. Among the most popular and widely utilized systems in use today are hepatic microsomes. These preparations retain activity of those enzymes which reside in the smooth endoplasmic reticulum, such as cytochromes P450 (CYP), flavin monooxygenases, and glucuronosyltransferases. Isolated hepatocytes appear to retain a broader spectrum of enzymatic activities, including not only reticular systems, but cytosolic and mitochondrial enzymes as well. Because of a rapid loss of hepatocyte-specific functions, it has been possible to generate useful data only with short term hepatocyte incubations or cultures [8]. However, significant strides have been made toward maximizing the viability of hepatocytes in culture. Liver slices, which like hepatocytes retain a wide array of enzyme activities, are also increasing in popularity. Furthermore, both hepatocytes and liver slices are capable of assessing of enzyme induction in vitro. The choice of which system to employ in a drug discovery screening program will depend on many factors, not the least of which is availability of tissue for large-scale implementation. In addition, historical information about a particular chemical series is invaluable. Consider, for example, an SAR-driven chemical series with a prototype analogue that has been demonstrated to induce CYP activity. For metabolic screening in this program, hepatocytes or liver slices may be preferable to microsomes. Biotransformation of NCE candidates as well as induction potential can be evaluated in the same system, making the added resources required to establish these systems well worth the investment. Information gathered from in vitro metabolism studies is especially useful in choosing drug candidates for future development. The potential utility of in vitro metabolism data in predicting in vivo intrinsic clearance has been touted [9,10].

Isolated heterologous human CYP enzymes have been available for several years, being expressed from cDNA in yeast (Saccharomyces cerevisiae), bacterial (Escherichia coli), and mammalian (B-lymphoblastoid) cell lines [11-13]. These systems can be used to ascertain whether a compound is a substrate for a particular CYP isozyme and, if so, what metabolite is generated by that enzyme. Moreover, these enzymes, in sufficient quantity, may possibly be used as bioreactors to generate usable amounts of a metabolic product that may be difficult to chemically synthesize in the laboratory. Isozyme-specific antibodies and isozyme-specific inhibitory substrates, by selectively abolishing the activity of a particular CYP isoform, may be used to determine the relative importance of that enzyme in the turnover of an NCE. Conversely, the ability of an NCE to interact with a particular CYP isozyme may be determined by metabolic cross-inhibition studies against prototypical CYP substrates.

In vitro metabolism systems are not limited to those derived from the liver. Most pharmaceutical companies have been building liver and tissue banks to permit a cross-comparison of metabolic turnover rates in various tissues from various species. It is not uncommon now for liver banks to house tissues from a variety of species, including those from animals treated with enzyme inducers or inhibitors. Therefore, cross-species in vitro metabolism comparisons are becoming more feasible and commonplace. In addition to providing information on potential rates and routes of metabolism, interspecies comparisons may help in choosing species to be used in toxicology studies.

Available in vitro metabolism technologies are considerably more efficient than traditional in vivo methods. With these systems, it is possible to assess the relative rates and routes of biotransformation of a handful of compounds in the time required to comparably characterize one compound using in vivo approaches. In the past, a relative paucity of tissues was the most notable impediment to conducting large-scale screening efforts with in vitro tools. However, the continued growth of in-house tissue banks, in concert with the optimization of isolated enzyme expression systems, have made high-throughput metabolism assessments in the discovery arena a reality.


Unfortunately, there is no substitute for actual in vivo data in assessing pharmacokinetic profiles of drug candidates. While insight into various aspects of the pharmacokinetic profile (absorption, metabolism, protein binding) can be gleaned from in vitro techniques, there are as yet no methods available for accurately predicting what will happen to a drug when it is put into a whole animal.

For a useful assessment of pharmacokinetics and bioavailability, it is necessary to administer the drug to selected animal species both intravenously and by the intended route of administration (usually oral). Blood samples are collected over a predetermined time course after dosing and the drug is quantified in serum or plasma by a suitable bioanalytical method (e.g. HPLC). Alternatively or concurrently it may be possible to collect plasma samples from animals used in whole animal pharmacologic models and, based on the concentration/effect (pharmacokinetic/ pharmacodynamic) relationship established, make a link between in vitro pharmacologic activity and the behavior of a compound in vivo [14].

In any case, the most significant impediment to providing pharmacokinetic input to a high-throughput discovery team is the time and labor-intensiveness of the bioanalytical methods available. To assess biochemical or pharmacologic activity in vitro or in vivo, a standardized screening method is established with a common assay endpoint that can be applied to test all compounds that are available. In contrast, a separate bioanalytical method for pharmacokinetic assessment must be developed for each compound. For chromatographic assays, which comprise the vast majority of the methods employed in pharmacokinetic studies, remarkable improvements in assay detection limits and sample cleanup have been realized. However, little has been done to shorten the time required for assay development and implementation. Pharmacokinetic characterization is therefore often relegated to end-stage discovery, being utilized to select which of 2 or 3 potential lead candidates has the most "acceptable" pharmacokinetic profile. Clearly, if pharmacokinetic input is to be available for early discovery decisions, more efficient methodologies are necessary.

Semi-simultaneous bioavailability estimation is a screening method which has been successfully utilized in drug discovery [15]. With this technique, an intravenous dose of drug is administered and at a suitable time postdose (i.e. postdistributional) an oral dose is administered to the same animal. Pharmacokinetic parameters such as clearance and volume of distribution can be determined from the intravenous concentration-time curve. Oral pharmacokinetic parameters, including bioavilability, are subsequently extricated from the combined IV and PO concentration-time data by deconvolution. By reducing inter- and intraanimal variability, accurate pharmacokinetic data can be gathered with fewer animals. In addition, the total number of plasma samples from both routes of administration is reduced by the overlapping dosing regimes. On the other hand, a prior knowledge of the pharmacokinetic characteristics of the NCE is needed to optimize the administration paradigm. In that regard, it may be possible to extrapolate from studies with structurally analogous predecessors from the same chemical series. More critical, however, is that compound-specific analytical methods are still required for each NCE.

Bioassays represent one alternative to compound-specific assays. Plasma samples are added to an in vitro system and the biological activity is quantified in much the same way as a pure drug solution is tested for activity. The extent of displacement of a prototype receptor ligand or inhibition of a target enzyme activity is measured as the assay endpoint for each sample. Based on standard curves which relate the assay endpoint to known quantities of added drug, the concentration of the compound in the plasma can be quantified. Microbiological assays, which measure the inhibition of bacterial growth on an agar plate in relation to the amount of antibacterial drug applied, are among the most widely utilized examples of this technique. The greatest advantage of bioassays over HPLC methods is that they can be employed for analyzing any compound which possesses the appropriate biological activity, making them amenable to high throughput screening. Unfortunately, owing to their non-specific nature, these assays will quantify all biological activity in a sample, which could include active metabolites. These results may be misleading if one is attempting to understand the pharmacokinetic behavior of only parent compound. In response, it could be argued that it is prudent to choose the compound which provides the best activity profile, irrespective of whether the in vivo activity observed is due to parent compound or active metabolite. Preprocessing of plasma samples (extraction) could possibly remove polar metabolites, allowing the bioassay to concentrate on parent drug.

In perhaps the most significant technologic leap in this field, scientists are capitalizing on the exquisite selectivity of mass spectrometry (MS) to circumvent the tedium of HPLC assay development. The resulting procedure is the pharmacokinetic equivalent of a mass-screen, the simplicity and elegance of which is the result of several evolutionary steps. In its earliest incarnation, animals were dosed with a single drug and samples were collected, processed, and injected onto an HPLC- MS. The chromatographic system was optimized to isolate the drug peak and the mass spectrometer was used to detect the appropriate mass ion and quantify it. Eventually, HPLC separation was foregone in favor of direct injection of a minimally preprocessed plasma sample into the MS. In a further refinement, multiple drugs were boldly administered simultaneously to the same animal and each compound was then individually quantified. MS is capable of distinguishing subtle differences in the characteristic mass ion peaks of each analyte, imparting a degree of specificity that HPLC cannot achieve without fairly elaborate intervention. Under this protocol, the number of drugs for which pharmacokinetic profiles can be obtained in a week increases several-fold, being limited only by the number of compounds which are co-administered in each animal. The considerable sample burden thus engendered is easily accommodated by the facile, direct-injection MS analysis. Overcoming myriad theoretical reasons why it shouldn't work (e.g. potential for metabolic interactions, protein binding displacement, or additive pharmacologic effects), investigators in several large pharmaceutical companies have successfully incorporated this screening approach into routine drug discovery support.

Drug distribution and tissue penetration data are occasionally obtained during discovery screening. Plasma protein binding assessments can be relatively easily performed using well-known and widely-used methods, providing perspective on the relationship between total in vivo plasma drug concentrations and those of pharmacologically available unbound drug. For more specific applications, cell cultures techniques are available for predicting the ability of compounds to traverse target biological membranes (e.g. blood-brain barrier, T lymphocytes). In vivo microdialysis has been used for assessing drug penetration into CNS and liver, but is methodologically difficult to transform into a screening tool. In addition, these methodologies typically suffer from the same bioanalytical limitations of classical pharmacokinetic profiling. It seems reasonable to hope that direct-injection MS analysis may facilitate these types of studies in the future.

ADME in Drug Design

With continued refinement and streamlining of its underlying technologies, ADME-relateddiscovery screening can be implemented ever earlier into the discovery process. Ultimately,this will permit scientists in the ADME disciplines to play a significant role inoptimizing new drug design. Fundamentally, that role is to identify the limiting barriers to optimal in vivo efficacy.

Of all the attributes which determine a drug's ultimate in vivo efficacy, physicochemical behavior is perhaps the most basic. For this reason alone it deserves considerable scrutiny by the drug discovery team. Moreover, as solubility, pKa, and lipophilicity are so integrally linked with chemical structure, physicochemical properties in some respects can be relatively predictably manipulated by chemical modifications. Historically, chemical synthetic efforts have been guided almost exclusively by SAR intended to maximize interaction with a therapeutic target (e.g. receptor, enzyme). The goal of ADME-guided synthesis, on the other hand, is to maximize the ability of an NCE to access the therapeutic target in vivo. As these synthetic directives may often be at odds, the Holy Grail for discovery teams in the future will be to achieve an optimal balance between the interests of these groups and fully integrate the valuable information which they can each impart toward guiding new drug design.

Lipophilicity is a critical determinant of a host of ADME processes. It has been suggested, for example, that compounds with log P values between 0 and 3 or 4 (e.g. 0.5 to 2) are the most suitable candidates for passive transcellular absorption across intestinal epithelia [2,16]. As lipophilicity is increased (log P greater than 3 or 4), solubility and hence absorption progressively decline. In contrast, more hydrophilic compounds with log P values less than 0 are likely to traverse the epithelium more slowly via paracellular channels, although it should be appreciated that molecular size is also an important determinant of transit through the narrow (5-9 Å) paracellular channels. Similarly, it has been observed that compounds with log P values greater than 0 are likely to undergo substantial renal tubular reabsorption. By itself, this effect will tend to prolong t1/2, however the greater lipophilicity and minimal renal excretion will also render the compound more susceptible to metabolism.

Experimental determination of lipophilicity is not a trivial matter, a fact which has limited the availability of this information in early discovery where it could be most useful. However, mathematical predictions based on a weighting of functional groups within a molecule have been performed with impressive accuracy [17]. As confidence in the reliability of these predictions is further developed, it will become more feasible in the synthesis of new drugs to target log P values to achieve a balance between particular ADME processes.

Even more so than log P, pKa is a physicochemical property that can be readily predicted, given the vast experience amassed over many years with myriad ionizable organic functional groups. The ionization state of an NCE in vivo can be predicted by considering its pKa relative to pH in various body compartments. It is possible to surmise, for example, that an organic acid will present to the nephron in an anionic state, rendering it potentially susceptible to active tubular secretion by the organic anion transporter. If the compound is indeed a substrate for this transporter, renal clearance may be high and elimination from the systemic circulation rapid.

Both log P and pKa are the primary determinants of compound solubility. In vivo, this has a decisive bearing on oral absorption, following the premise that drugs must dissolve in the GI tract to be available to cross the intestinal membrane. Given that pH progressively increases throughout the length of the gastrointestinal tract, solubility and thus absorbability of an ionizable compound will vary accordingly. Solubility, log P, and pKa are closely intertwined determinants of renal tubular reabsorption as well. Thus, whereas compounds with log P values greater than 0 may be potential candidates for tubular reabsorption, this process will only be likely to occur if the compound is present in urine (pH 4.5 to 8) in an un-ionized state and has not exceeded its solubility limit.

Despite the importance of physicochemical properties in determining the ADME profile, they are as yet primarily rough guidelines and cannot substitute for experimentation to bear out the validity of their predicted in vivo effects. Experimental data generated by the discovery tools outlined above can help put the importance of physicochemical properties into the appropriate context.

Structural modifications may also be guided by more fundamental knowledge of a particular ADME process. The stability of peptidomimetic compounds to peptidases in the GI tract may be improved by various synthetic alterations, including protection of peptide bonds with steric bulk or substitution of naturally occurring L-amino acid residues with their D-isomers. The increased GI stability potentially increases bioavailability, but must be balanced against the contribution of physicochemical properties such as lipophilicity [18,19]. In addition, it is important to note that although lipophilicity and GI stability may favor intestinal absorption, high molecular weight (>600 Da) of some of these compounds could render them acutely sensitive to biliary excretion, reducing absolute systemic bioavailability in spite of their good intestinal permeability.

Among the many transport proteins which reside in the GI mucosa, the dipeptide transporter has been shown to be an important mediator of active drug absorption. In addition to di- and tripeptides, compounds from several compound classes which are structurally similar to di- and tripeptides have also been shown to be substrates for this carrier, including -lactams and ACE inhibitors. Targeting synthesis to enhance the affinity of NCEs for this transporter may prove to be a fruitful approach to improving systemic bioavailability, particularly for polar compounds with inherently low membrane permeabilities. In one such example, formation of a proline adduct improved the transcellular passage of L-methyldopa by transforming it into a substrate for the dipeptide transporter [20]. In addition to being dipeptide-like, L-methyldopa-pro is also a substrate for intracellular prolidase which was specifically exploited to regenerate L-methyldopa after the adduct was transported across the apical surface of the enterocyte. As their substrate specificities are further delineated, other known transporters could be similarly targeted for enhancing bioavailability.

Prodrug strategies have successfully improved the oral bioavailability of numerous compounds. In many cases, this involves masking a polar group by esterification to increase lipophilicity and enhance the extent of absorption from the GI tract. After absorption, the ester is enzymatically hydrolyzed to release parent drug. Absorption has also been improved with prodrugs designed to enhance solubility by decreasing lipophilicity, targeting progroup cleavage for the GI tract. With this approach, the nature of the progroup predetermines whether parent drug is released in the GI lumen or at the brush border membrane [21]. The ideal candidate for the latter manipulation is a highly lipophilic compound which readily traverses the GI mucosa, but is too insoluble to produce a reasonable soluble fraction for acceptable bioavailability. A polar progroup (e.g. phosphate, amino acid) is incorporated into the molecule, markedly increasing its solubility in the GI tract. The progroup is subsequently cleaved at the intestinal brush border by the targeted enzymes (e.g. phosphatases, aminopeptidases), which expels the lipophilic parent compound in close proximity to the intestinal surface. The low local concentration minimizes the possibility of precipitation and allows the drug to be directly absorbed. The bioavailability of several compounds has been improved with this strategy [22,23].

Soft drugs provide an interesting counterpoint to prodrugs. In both cases, the administered compound is designed to undergo a predictable metabolic modification. With prodrugs, this biotransformation event rapidly converts an inactive precursor into an active drug. Conversely, a soft drug is itself the active drug which, after eliciting a therapeutic effect, is metabolically inactivated in a predictable manner and at a controlled rate. Several approaches to rational soft drug design have been identified, including soft analogs, activated soft compounds, natural soft drugs, the active metabolite approach, and the inactive metabolite approach [24]. In the latter case, an inactive metabolite to a known drug is first identified. This metabolite is then chemically modified to create a new drug which:

  1. possesses comparable pharmacologic activity to the original drug,
  2. will be biotransformed to the inactive metabolite in one step and without going through toxic intermediates, and
  3. may possibly be subjected to further modification if transport or binding properties require optimization.

The metabolic systems typically harnessed by soft drugs are hydrolytic enzymes (e.g. esterases). Monoxygenases are scrupulously avoided in the belief that hydrolyses are inestimably more predictable and controllable processes. In reality, it may be difficult to integrate the soft-drug concept into novel, innovative drug discovery programs. Its retrometabolic philosophy is based on the premise that the metabolic fate of a pharmacologically active drug is known in explicit detail. Therefore, unless the synthetic strategy is predicated on a well-established chemical series, the information required for soft-drug design will not be available. However, the basic principles of this concept might possibly be extrapolated from successful soft drugs to novel compounds with similar functional moieties.

The guidance of drug design based on more classical metabolism theory is multifaceted. As with any strategy, an ability to predict the metabolic fate of an NCE based simply on its structure would be of untold value. To that end, several computerized " Expert systems" have been developed which are designed to promulgate for any given compound a list of likely metabolic products. Unfortunately, these systems have proven to be perhaps too general in their scope, projecting an array of potential products so extensive as to be of relatively limited use. On a slightly different tack, a vast metabolite database is currently being developed which is based on metabolism studies reported in the literature over the past 90 years. This package, from Molecular Design [25], interfaces with ISIS databases and is thought to hold considerable promise for providing more realistic projections of metabolic routes across species.

Molecular modeling may come to fruition in the near future. Computer assisted drug design might allow the determination of NCE-receptor compatibility to proceed concurrently with an evaluation of the ability of the NCE to fit into the active site of a particular metabolic enzyme. This capacity does not presently exist on any significant scale. It is particularly frustrating with respect to the CYP family, which is the most important of Phase 1 metabolic enzyme systems. In part because CYP is a membrane-bound enzyme in close association with a reductase, it has not yet been crystallized. Therefore, an X-ray structure, required for building a useful computer model, is not available. The multiplicity of the CYP family makes the task even more formidable.

Therefore, guiding synthesis toward an optimal metabolic profile is based as yet on largely empirical insight. Substantial inroads have been made in this regard, particularly with respect to 4 of the most important CYP isozymes, CYP2D6, 2C9, 2C19, and 3A4. Models of the active sites of CYP2D6 and CYP2C19 have been hypothesized based on a comparison of chemical structures of a variety of substrates, taking into consideration the strict stereo- and regioselectivity of their transformations. The high degree of selectivity inherent in their active sites make it feasible to design drugs which may be more or less likely to interact with these isozymes. In one example, addition of a bulky para-substituent to the basic beta-blocker nucleus ultimately yielded betaxolol, which unlike its pharmacologic brethren is resistant to the action of CYP2D6 [16]. CYP3A4, on the other hand, has to date defied attempts to define its active site, being far less discriminating in its choice of substrates. Indeed, some rather large molecules, such as cyclosporin and macrolide antibiotics are known to be turned over by CYP3A4. As metabolism databases expand in the coming years, a better delineation of how to chemically modify NCEs will be forthcoming, permitting more predictable interactions with CYP isozymes.

Some general rules are in place with regard to avoiding chemically reactive structures. Thus, compounds containing structural elements with a high risk of producing toxic products should be strictly avoided. Examples include quinones and quinonimines, aromatic methylene-dioxy groups, and aromatic nitro groups [26].

Enzyme induction is more difficult to predict. There is little guidance with regard to what chemical structures are known to produce this effect, with the possible exception of broad chemical groups such as barbiturates, hydantoins, di-t-butyl phenols, etc. However, even within such series, subtle changes in chemical structure can markedly and unpredictably alter enzyme induction potential [27].

As detailed above, pharmacokinetic characterization is often relegated to end- stage discovery, at which point compound attrition has whittled the pool of potential lead compounds to a small, manageable number. Under these circumstances, the opportunity to provide meaningful feedback which might assist in pharmacokinetically optimizing drug design is limited. As analytical techniques become more expeditious, pharmacokineticists will play an increasingly important role in drug design. Nonetheless, it should be clearly recognized that pharmacokinetics in and of itself is not particularly meaningful. Pharmacokinetic behavior is highly interdependent with absorption and metabolism. However, characterizing the pharmacokinetic profile of a compound is still of utmost importance insofar as it helps put absorption, metabolism, and other data into the proper perspective. For example, in vivo activity may be poor despite predictions from in vitro absorption and metabolism studies that an NCE will be well absorbed and not be subjected to appreciable biotransformation. The pharmacokinetic profile reveals that the drug has a short in vivo half-life owing to a small volume of distribution and high renal clearance, neither of which could have been predicted accurately prior to the whole animal work. In this case, an optimization of the PK profile may be guided by a realization that chemical modifications are needed which will specifically reduce renal clearance or alter plasma protein binding.

The ability to predict a priori the extent of plasma protein binding is limited, although fruitful attempts have been made to define the molecular binding requirements for certain sites on human serum albumin. In general, binding is likely to increase with lipophilicity; an anionic group may enhance binding to albumin even further [28,29]. In contrast, cationic molecules are virtually precluded from binding to albumin, rather favoring -1-acid glycoprotein. Ultimately, despite formidable theoretical arguments to the contrary, the role of protein binding in determining in vivo efficacy is still not universally predictable. For example, it has been suggested that NCEs with low protein binding are the most ideal candidates for further development [26]. Quite often this implies low lipophilicity, which could be at odds with strategies for optimizing membrane permeability or receptor or enzyme affinity. It has also been suggested that high protein binding is not necessarily a negative attribute and in fact may help prolong the activity of some compounds with otherwise short half-lives [30].


By now the advantages of ADME input into drug design should be obvious. A full consideration of all facets of the molecular structure and the impact of that structure on the in vivo ADME profile will enable chemists to design out negative ADME attributes (e.g. chemically reactive moieties, avidly metabolized sites) and incorporate "ADME-friendly" attributes (e.g. optimal log P, good membrane permeability).

It may not be possible to achieve the ideal ADME profile. However, this shouldn't necessarily preclude further development of a lead compound. For example, if a subtle chemical alteration designed to improve the ADME profile has been shown to sacrifice otherwise exquisite pharmacologic activity, it would be inadvisable to incorporate that alteration. In this case, although SAR concerns may appear to have outweighed ADME concerns, the optimal SAR/ADME balance has in fact been achieved.

New screening technologies will allow the incorporation of ADME profiling into the earliest phase of the discovery process, which should facilitate development of lead compounds even in the absence of optimized drug design. Familiarity with the compound afforded by early and sustained involvement in its discovery helps provide substantial fundamental information about its profile. Thus, numerous obstacles which would otherwise retard the development process can be foreseen and dealt with proactively rather than reactively. From an ADME perspective, a glimpse into the pharmacokinetic and metabolic profile prior to initiating development work is of untold value in designing toxicology/toxicokinetic and pharmacokinetic studies to support an IND.


1. Cayen MN. Considerations in the design of toxicokinetic programs. Tox Path, 1995; 23: 148-157.

2. Austel V. The medicinal chemist's approach, in Modern Drug Research, Paths to Better and Safer Drugs. Martin YC, Kutter E, Austel V, Eds, Marcel Dekker, Inc., New York, 1989, pp 243-307.

3. Inui K-I, Yamamoto M, Saito H. Transepithelial transport of oral cephalosporins by monolayers of intestinal epithelial cell line CaCO-2: specific transport systems in apical and basolateral membranes. J Pharmacol Exp Ther, 1992; 261: 195-201.

4. Lu S, Guttendorf RJ, Stewart BH. The mechanism of cefdinir transport in CaCO-2 cells. Pharm Res, 1994; 11: S-258.

5. Bjorge S, Halelehle KL, Homan R, Rose SE, Turluck DA, Wright DS. Evidence for glucuronide conjugation of p-nitrophenol in the CaCO-2 cell model. Pharm Res, 1991; 8: 1441-1443.

6. Stewart BH, Chan OH, Lu RH, Reyner EL, Schmid HL, Hamilton HW, Steinbaugh BA, Taylor MD. Comparison of intestinal permeabilities determined in multiple in vitro and in situ models: relationship to absorption in humans. Pharm. Res., 1995, 12: 693-699.

7. Pidgeon C, Ong S. Predicting drug-membrane interactions. Chemtech, June 1995: 38-48.

8. Hawksworth GM. Advantages and disadvantages of using human cells for pharmacological and toxicological studies. Hum Exp Tox, 1994; 13: 568-573.

9. Hoener B-A. Predicting the hepatic clearance of xenobiotics in humans from in vitro data. Biopharm Drug Disp 1994; 15: 295-304.

10. Houston JB. Utility of in vitro drug metabolism data in predicting in vivo metabolic clearance. Biochem Pharmacol, 1994; 47: 1469-1479.

11. Ohgiya S, Komori M, Fujitani T, Miura T, Shinriki N, Kamataki T. Cloning of human cytochome P-450 cDNA and its expression in Saccharomyces cerevisiae. Biochem Int, 1989; 18: 429-438.

12. Winters DK, Cederbaum AI. Expression of a catalytically active human cytochrome P-4502E1 in Escherichia coli. Biochim Biophys Acta 1992; 1156: 43-49.

13. Crespi CL, Gonzalez FJ, Steimel DT, Turner TR, Gelboin HV, Penman BW, Langenbach R. A metabolically competent human cell line expressing five cDNAs encoding procarcinogen-activating enzymes: aplication to mutagenicity testing. Chem Res Toxicol, 1991; 4: 566-572.

14. Guttendorf RJ, Andress L, Okonkwo G, Lesch M, Finkel M, Schrier D, Loewen G. Pharmacokinetic and pharmacodynamic assessment of a novel antiinflammatory drug in early discovery. Pharm Res, 1992; 9: S- 356.

15. Karlsson MO, Bredberg U. Estimation of bioavailability on a single occasion after semisimultaneous drug administration. Pharm Res, 1989; 6: 817-821.

16. Smith DA. Design of drugs through a consideration of drug metabolism and pharmacokinetics. Eur J Drug Metab Pharm, 1994; 3: 193-199.

17. Moriguchi I, Shuichi H, Liu Q, Nakagome I, Matsushita Y. Simple method of calculating octanol/water partition coefficient. Chem Pharm Bull, 1992; 40: 127-130.

18. Humphrey MJ, Smith DA. Role of metabolism and pharmacokinetic studies in the discovery of new drugs - present and future perspectives. Xenobiotica 1992; 7: 743-755.

19. Hamilton HW, Steinbaugh BA, Stewart BH, Chan OH, Schmid HL, Schroeder R, Ryan MJ, Keiser J, Taylor MD, Blankey CJ, Kaltenbronn JS, Wright J, Hicks J. Evaluation of physicochemical parameters important to the oral bioavailability of peptide-like compounds: implications for the synthesis of renin inhibitors. J Med Chem, 1995; 38: 1446-1455.

20. Bai JP-F, Hu M, Subramanian P, Mosberg HI, Amidon GL. Utilization of peptide carrier system to improve intestinal absorption: targeting prolidase as a prodrug-converting enzyme. J Pharm Sci, 1992; 81: 113-116.

21. Fleisher D, Stewart BH, Amidon GL. Design of prodrugs for improved gastrointestinal absorption by intestinal enzyme targeting. Meth Enzymol 1985; 112: 360-381.

22. Amidon GL, Stewart BH, Pogany S. Improving the intestinal mucosal cell uptake of water insoluble compounds. J Cont Rel, 1985; 2: 13-26.

23. Fleisher D, Johnson KC, Stewart BH, Amidon GL. Oral absorption of 21-corticosteroid esters: a function of aqueous stabiliity and intestinal enzyme activity and distribution. J Pharm Sci, 1986; 75: 934-939.

24. Bodor N. Soft drugs: principles and methods for the design of safe drugs. Med Res Rev, 1984; 4: 449-469.

25. Buchalter C. Formula searching in ISIS/Base. Scientist's desktop. A newsletter for ISIS and CPSS Scientific software users, 3(1). Spangler S, Ed., MDL Information Systems Inc., San Leandro, CA, 1994.

26. Bozler G, Schmid J. Principles of pharmacokinetics and drug metabolism, in Modern Drug Research, Paths to Better and Safer Drugs. Martin YC, Kutter E, Austel V, Eds, Marcel Dekker, Inc., New York, 1989, pp 77-160.

27. Loewen G, Patel B, Buckley K, Connor D, Mullican M. Unangst P. Structure-activity relationship of enzyme inductino in beagle dogs within a series of antinflammatory di-t-butyl phenols. Pharm Res, 1992; 9: S-296.

28. Sudlow G, Birkett DJ, Wade DN. The characterization of two specific drug binding sites on human serum albumin. Mol Pharmacol, 1975; 11: 824-832.

29. Wanwimolruk S, Birkett DJ, Brooks PM. Structural requirements for drug binding to site II on human serum albumin. Mol Pharmacol, 1983; 24: 458-463.

30. McNamara PJ. Protein binding - is it relevant in clinical practice? Insights into the Treatment of Serious Infections, 1988; 2: 30-37.

Related Web Sites:

American Association of Pharmaceutical Scientists (AAPS): (

World-Wide Web Virtual Library: Pharmacy (Miscellaneous Pharmaceutical Resources):

NetSci, ISSN 1092-7360, is published by Network Science Corporation. Except where expressly stated, content at this site is copyright (© 1995 - 2010) by Network Science Corporation and is for your personal use only. No redistribution is allowed without written permission from Network Science Corporation. This web site is managed by:

Network Science Corporation
4411 Connecticut Avenue NW, STE 514
Washington, DC 20008
Tel: (828) 817-9811
Website Hosted by Total Choice