Frontiers in Medicinal Chemistry: Volume 8

PREFACE

Volume 8 of Frontiers in Medicinal Chemistry comprises 6 chapters on topics of high importance in the fields of medicinal chemistry and early-stage drug discovery research. The topics and authors were selected from highly cited publications in the Bentham journals Curr. Med. Chem., Curr. Pharm. Des., and Curr. Top. Med. Chem. The original authors were given the opportunity to rewrite their contributions, particularly updating them with more modern insights and references that have emerged in the intervening period of time.

The first chapter by Dr. Gary Caldwell is a tour de force review on the use of ADME optimization and toxicity assessment in drug discovery research. Dr. Caldwell has written multiple chapters on this ever evolving topic and has served as a book editor for Bentham in the past. His chapter in this volume of Frontiers is a valued and updated reference guide in the field. Chapters 2 and 3 by Briasoulis et al. and Escribano-Lopez et al., respectively, cover oxidative stress-mediated approaches to treat cardiovascular diseases such as atherosclerosis. The fourth chapter by Potenza et al. is on diabetes, and in particular endothelial dysfunction and the associated implications as to mechanism and therapeutic targets. In Chapter 5, Jokanović and Petrović review pyridinium oximes as cholinesterase reactivators for the treatment of organophosphorus poisoning. Finally, Rodik and colleagues discuss the current and potential uses of the cyclic macrocyclic oligomer calixarenes in biomedical research.

I would like to express my gratitude to all the authors for their excellent contributions. I would also like to thank the entire team of Bentham Science Publishers, particularly Mr. Omer Shafi (Assistant Manager Publications), Mr. Shehzad Naqvi (Senior Manager Publications) and team leader Mr. Mahmood Alam (Director Publications) for their excellent efforts. We are confident that this volume will receive wide appreciation from students and researchers.

ADME Optimization and Toxicity Assessment in Drug Discovery

Gary W. Caldwell*

Janssen Research & Development, L.L.C. Discovery Sciences, Lead Discovery Department, Welsh and McKean Roads, P.O. Box 776, Spring House, PA 19477, USA

Abstract

Integrating physicochemical, absorption, distribution, metabolism, excretion, pharmacokinetics, and toxicity data into the drug discovery/preclinical development decision process in order to reduce the attrition rates of new chemical entities (NCEs) in clinical development is reviewed. The review is organized around the three main stage gates in a small molecule target-based approach including hit-to-lead (H2L), lead optimization (LO) and the final stage gate for selecting NCEs for entry into Phase I clinical trials. The preclinical in silico computational methods and in vitro cellular assays utilized at each stage gate are discussed from a drug discovery perspective. Preclinical assays utilized at the H2L and LO stage gates must have turn-around-times within a timeframe that is consistent with the iterative cycle of the research projects and consume small quantities of compounds while at the final NCE stage gate more traditional assays are used. Unfortunately, many preclinical assays are ambiguous in predicting human preclinical data since they contain a significant amount of false-positive and false-negative information and, therefore, are not easily translatable from cellular/animals to humans. Thus, understanding the limitations of these preclinical assays is a must for all medicinal chemists for developing go/no-go selection criteria and drug-design optimization strategies to advance small molecule drug candidates through the various stage gates of a target-based screening approach.

Keywords: ADME, blood brain barrier, cell toxicity, CYP450, DNA binding, efflux, formulation, hERG, hit-to-lead, in silico computational methods, in vitro assays, in vivo animal studies, induction, inhibition, Irwin test, lead optimization, metabolic stability, mutagenicity, new chemical entity, oral bioavailability, oxidative stress, permeability, pharmaceutical industry, physicochemical, protein binding, reactive intermediates, screening, solubility, target selection, target-based screening.

* Corresponding author Gray W. Caldwell: Janssen Research & Development, L.L.C. Discovery Sciences, Lead Discovery Department, Welsh and McKean Roads, P.O. Box 776, Spring House, PA 19477, USA; Tel: 215-628-5537; Fax: 215-628-7064; E-mail: [email protected]

INTRODUCTION

Moving forward in the 21st century, the goal of the pharmaceutical industry continues to be the discovery and delivery of life-saving medicines (i.e., drugs) that improves global public health care. The pharmaceutical process that produces drugs for patients can be broadly categorized into a discovery/preclinical development step, a clinical drug development step, and a commercialization step (Figs. 1 and 2). Each step in the process works somewhat independently from each other; however, the entire end-to-end pharmaceutical process has overarching common goals for each of the three steps to minimize timelines and financial investments [1]. The drug discovery/preclinical development step in a small molecule target-based screening approach is managed by subdividing the discovery/preclinical development step into stage gate research processes such as target selection, high-throughput screening (HTS), hit-to-lead (H2L), lead optimization (LO), and new chemical entity (NCE) selection. The stage gates utilize compound selection criteria based primarily on in silico computational methods, in vitro assays, and in vivo animal models to advance drug candidates from H2L to the LO and finally to the selection of an NCE. Thus, the main goal of the drug discovery/preclinical development step is to select NCEs that have a high probability of becoming a safe drug after receiving more extensive and time-consuming clinical testing. The clinical development step is subdivided into human safety clinical trials (Phase I), and drug efficacy clinical trials (Phase II, and III). The commercialization step involves the development of marketing strategies that ensures physicians can quickly and safely prescribe new drugs to patients. In some cases, post-marketing surveillance trials (i.e., Phase IV) are conducted to monitor long-term effectiveness and impact of the drug on the quality of life of patients. In the US, the movement of NCEs from the drug discovery/preclinical development step to clinical trial steps requires the filing of an Investigational New Drug application (IND) with the US Food and Drug Administration (FDA). The movement of drugs from clinical trials to commercialization requires the filing of a New Drug Application (NDA) followed by final FDA approval of the drug to a new molecular entity (NME) status [2]. The time to complete the entire pharmaceutical process for small molecule drug candidates to become NMEs - on the average - ranges from 10 to 15 years with the discovery/preclinical development step accounting for approximately 3 to 5 years and the clinical development and commercialization steps accounting for 7 to 10 years [3-5]. The cost to produce a single NME has increased steadily over the years with financial investments today ranging from 1 to 2 billion dollars [6]. About one-third of total expenditures for the pharmaceutical process is spent on the discovery/preclinical development step and two-thirds on the clinical development and commercialization steps. The average success rate for the pharmaceutical industry to discover and market a small molecule drug - for all therapeutic areas - has been estimated to range from 7% to 11% [7]. Therefore, while the pharmaceutical process to discover and deliver drugs to patients has an underlying scientific industrialization structure, the success rate is very low primarily due to incomplete knowledge of disease pathology and pharmacology with serendipity continuing to play a significant role in the process.

The FDA has approved approximately 1,408 small molecule NMEs between the years 1950 to 2014 [8]. On the average, the pharmaceutical process has produced 20 to 25 small molecule NMEs annually for the past 60 years and this average value has not trended upward as expected based on advancements in the understanding of disease pharmacology, chemistry, biology, informatics, computational and analytical technologies [4-6]. This absence of sustained productivity, significant attrition rates, government regulatory regulations, and rising costs for NMEs - particularly within the last 25 years - has generated great concern within the pharmaceutical industry as evident by the many publications that have analyzed the metrics and trends that drive research and development (R&D) productivity [9-28]. The R&D productivity metric and trends suggest that the high attrition rates of NCEs in clinical development tend to be governed by unacceptable safety (Phase I and II) - possibly due to off-target effects or failure of animal models to translate to humans, lack of efficacy (Phase II) - possibly due to limiting target engagement issues or selecting targets that do not significantly influence the disease pharmacology, and economic reasons (Phase I and II) – possibly due to budget/resource constraints or a change in the company’s portfolio [20-26, 29]. It should be kept in mind that it is an oversimplification to assume that a single scientific factor governs the attrition rates of NCEs in clinical trials. It is more likely that attrition rates are governed by multiple factors since drug efficacy and safety deficiencies are related in part to pharmacokinetics (PK), toxicokinetics (TK), and drug-drug interactions (DDI), which are related in part to absorption, distribution, metabolism and excretion (ADME) properties of the drug [30].

For the past 25 years, thousands of scientists have conducted research to design surrogate in silico computational methods, in vitro ADME, and in vitro toxicity assays to predict human in vivo PK/efficacy and TK/safety at the discovery/ preclinical development step [31-100]. Therefore, most pharmaceutical companies today use panels of well-characterized in silico computational methods, in vitro ADME, and in vitro toxicity screens in parallel with in vivo animal PK, TK, efficacy and safety assays to identify drug candidates that have the potential of becoming NMEs at the discovery/preclinical development step (Figs. 1 and 2) [52, 56, 61, 64, 67, 72, 83, 99]. It is interesting to note that during the early 1990s, approximately 40% of NCE failures in Phase I clinical trials were attributed to ADME, and/or PK defects [9-11, 21]. After pharmaceutical companies started using panels of well-characterized ADME/PK screens to make earlier and smarter go/no-go decisions for drug candidates, the attrition rates of NCEs due to ADME/PK failures in Phase I clinical trials declined to approximately 11% in early 2000s [14, 21, 29]. While an 11% ADME/PK attrition rate is still significant, it highlights the fact that using in silico/in vitro/in vivo screens to uncover and correct ADME/PK defects in drug candidates early in the drug discovery process can be accomplished; however, correcting only ADME/PK defects is not enough to significantly increase the overall R&D productivity of NMEs. It suggests that the reason for the lack of increase in R&D productivity seen today stems from target validation issues and/or the failure of efficacy/safety in silico computational methods, in vitro cellular assays, and in vivo animal models to translate accurately to in vivo human predictions [3-5, 100].

The aim of this review is to provide a selected overview - based on the author’s experience - of in silico computational methods, in vitro ADME assays, and in vitro toxicity strategies that have been successful or unsuccessful from a drug discovery/preclinical development perspective [75, 99, 100]. Pharmaceutical scientists have recognized for many years that unacceptable physicochemical (LogP, pKa, solubility, etc.), ADME, PK, and toxicity properties are serious problems in advancing drug candidates to NCE status. It has been recognized that many of these pharmaceutical defects can be addressed at different stage gates during the drug discovery process [56, 64, 65, 81, 93, 95]. Since different types of preclinical activities take place at each of these difference stage gates, it is important to distinguish these preclinical activities and to recognize their advantages and disadvantages in designing selection criteria and drug-design optimization strategies. Therefore, we will address the following questions:

What in silico/in vitro/in vivo ADME and toxicity assays do we use in early- and late-phase drug discovery?

What are the advantages or disadvantages of these assays?

What types of decisions can be made using these assays?

Using the drug discovery concept outlined in Figs. (1 and 2 to select drug candidates for clinical development, we will focus on these questions for the H2L, LO and finally the NCE selection stage gates. Understanding the limitations of preclinical assays is important for all medicinal chemists in establishing selection criteria and drug-design optimization strategies to advance small molecule drug candidates through the drug discovery/preclinical development step.

Figure 1)

Preclinical drug discovery assays for physicochemical, ADME, and PK endpoints. See text for an explanation of the figure.

Figure 2)

Preclinical drug discovery assays for toxicity endpoints. See text for an explanation of the figure.

DRUG DISCOVERY/PRECLINICAL DEVELOPMENT

Since the drug discovery/preclinical development step of the pharmaceutical process is relatively inexpensive compared to the clinical development and commercialization steps, eliminating flawed drug candidates at this step is important since it will reduce overall R&D costs. The drug discovery/preclinical development step for a small molecule target–based screening approach starts with the selection of appropriate therapeutic targets (Figs. 1 and 2) [22]. A therapeutic target is defined as a particular biological target (i.e., enzyme, receptor, ion channel, etc.) that is hypothesized to be linked to the disease pharmacology. The target selection process is the starting point of all drug discovery programs and is highly influenced or even mandated by company franchises since it has to occur a decade or longer before the launch of an NME. In a target-based screening approach, target selection is typically a two-step process involving target identification and validation of its link to the disease pharmacology [101-103]. During the last decade, genomics, proteomics, and metabolomics approaches have drastically improved the target identification step [104]; however, the validation step may require many years of research to understand fully the disease pharmacology and thus, target validation remains the bottleneck in the process. Typically, the first stage gate, after target selection, is high throughput screening (HTS) of a primary binding-assay [105-108] using corporate chemical libraries to identify promising compounds that bind to the target [109-111]. The goal here is to screen large compound libraries in a relatively short amount of time attempting to find diverse compounds or families of structurally related compounds with significant potency against the selected target (i.e., hits). These hit compounds are clustered together based on similar structural motifs and ranked based on their potency to identify subsets of chemotypes. Hit compounds are filtered with substructure database searching algorithms (i.e., CAS SciFinder or PubChem) using similarity metrics to remove known drugs, compounds that contain reactive moieties (i.e., aldehydes, alkyl halides, Michael acceptors and so on), and promiscuous compounds that are known to inhibit many types of protein-targets [112]. There is a variety of in silico computational methods that can be used to filter hit compounds to identify and possibly to eliminate hits with potentially poor physicochemical, ADME and toxicity properties [100]. Once the hit or clustered chemotype have been validated, using various secondary target-related screens to eliminate false-positives and to confirm concentration-dependent potency activity, H2L compound series are declared. The H2L stage gate in combination with the LO stage gate drives the NCE drug candidate selection process. There are a variety of in silico and in vitro physicochemical, ADME and toxicity assays that can be used to guide the selection criteria and the drug-design optimization strategies to advance small molecule drug candidates through the H2L and LO stage gates. From these data, several drug candidates are selected as potential NCEs and will be screened to improve pharmacodynamic (PD) and PK/TK properties against the selected target, to establish efficacy and to eliminate any toxicity events using a range of in vitro assays and in vivo animal models. The overall goal of the final stage gate involves the selection of an NCE and collecting all necessary preclinical data for an IND submission of the NCE.

DRUG DISCOVERY PRECLINICAL SCREENS

Drug discovery preclinical screens must have turn-around-times within a timeframe that is consistent with the iterative cycle of drug discovery research projects. In addition, the data generated from these screens must have high enough quality to make the necessary go/no-go decisions for the advancement of drug candidates through the various stage gates within a target-based screening approach. Since no single factor can account for the NCE failures observed in clinical development, it is necessary to select a range of full and abbreviated ADME/PK and toxicity in silico/in vitro/in vivo assays that can provide maximum information in the discovery phase. Pharmaceutical companies use panels of well-characterized ADME/PK and toxicity screens to make better and earlier preclinical go/no-go decisions for drug candidates [52, 56, 61, 64, 67, 72, 75, 83, 99]. A scheme based on this drug discovery concept is shown in Figs. (1 and 2. Different types of assays are applied to each stage gate of the drug discovery/preclinical development step; that is, in our example, at the screening and H2L stage gate, the drug LO stage gate, and finally the NCE selection stage gate. At the screening and H2L stage gate, one needs to use assays that have sample throughputs that are relatively high since thousands of compounds can be selected at this stage of the process. Typically, in silico computational methods or ultra-fast in vitro assays - using compound library dimethylsulfoxide (DMSO) stock solutions - are used to sort compound properties into a high/low classification scheme. At the drug LO stage gate, one needs to use fast assays that have sample throughputs in the hundreds per day and consume small amounts of compounds in the range of 1 to 2 milligrams. Fast assays are defined as those that have turn-around-times within a timeframe that is consistent with the iterative cycle of drug discovery research projects. Using these types of fast in vitro assays, typically one would like to prioritize compounds from high-to-low to establish structure-activity relationships. The drug LO stage gate assays include physicochemical assays, abbreviated formulation screens, abbreviated in vivo PK assays, in vitro ADME, efflux, and P450 interactions (inhibition and induction) assays, to name a few. In the final selection stage gate, a few drug candidates are tested in assays requiring gram amounts of compounds for full in vivo PK and TK studies, acute and chronic dosing studies, drug-drug interaction, drug-serum protein binding, receptor (or enzyme) selectivity screening, and secondary genetic toxicity testing, safety pharmacology assays such as cardiovascular effects (CV) and central nervous system effects (CNS), developmental and reproductive toxicity testing, and clinical pathology.

Primary assays such as ultra-fast and fast in vitro or abbreviated in vivo assays should have reasonable sample throughput, require minimal amounts of compound and are validated; that is, the strengths and the weaknesses of each assay are fully evaluated to understand the false positives and false negative rates. For a drug candidate to pass a primary assay, the selection criteria must be clearly understood and established for each assay. It should be understood if the results from the assays could be prioritized from high-to-low or simply classified into a high/low classification scheme. The accuracy and precision of each assay need to be understood to establish selection criteria. Due to biological variability, the acceptable degree of accuracy for some assays might be as high as 10-fold with the precision, as measured by the percent coefficient of variation (% CV), ranging from 30% to 40%. Also, the precision may be more variable due to lab-to-lab biological sample variation. While there are many fast in vitro or abbreviated in vivo assays that are currently used in drug discovery, there are very few ultra-fast in vitro assays [75, 99]. The traditional in silico and in vitro physicochemical/ADME/toxicity assays utilized in H2L, LO, and NCE selection stage gates will be reviewed in the following sections.

H2L STAGE GATE (EARLY DRUG DISCOVERY)

For the past 25 years, due to the development of HTS in discovery, medicinal chemists have been faced with the daunting challenge of selecting good LO chemical starting points from thousands of possible starting points (i.e., hits). A significant amount of effort in the pharmaceutical industry and academic labs has been devoted to developing computer-based in silico approaches to calculate various physicochemical, ADME, PK and toxicity properties of compounds directly from their structures [113-130]. Having this type of data helps to prioritize hits not only based on their potency against the target but also based on their druggability [40, 113]. In addition, in silico approaches have the capability to detect and identify problematic structural motifs before they are synthesized, and also have the potential to rapidly screen virtual libraries of drug candidates to discover new hits [131-136]. There are many in silico tools in different combinations that can be used to establish selection criteria at the H2L stage gate [100]. In fact, there are so many different in silico techniques that have been developed in the last 15 years, many medicinal and computational chemists struggle to select those techniques that are beneficial for designing new compounds. Here, we will review only a few selected in silico approaches and examine their strengths and weaknesses.

In Silico Physicochemical Properties: LogP, LogD, pKa, and S0

Properties of virtual compounds are, in many cases, calculated based on the two-dimensional (2D) structure of the compound and simple empirical-based in silico models. Empirical-based models involve summing the contributions of structural units (i.e., atoms or fragments) within a compound to predict quantitatively the properties of the compound as a whole. The values of the structural units are derived from experimental data of known compounds for each property of interest. Once the structural units are known for a particular property, they can be combined in different ways to create real or virtual compounds and thus, by summing the value of the structural units in the compound its property is calculated. Structural properties of compounds, using an empirical-based approach, includes molecular weight (MW), hydrogen bonding characteristics, flexibility index, and molecular polar surface area, to name a few. These types of properties are easily calculated using the 2D structure of compounds and structural unit values of the property of interest [40, 113].

Physicochemical properties including lipophilicity (LogP or LogD), acid ionization ((pKa = -Log [H+]), and intrinsic solubility (S0) which are the most important properties of hits are also calculated from empirical-based 2D structural units [137-142]. Many reasonable in silico empirical-based models for calculating LogP, LogD, and pKa values have been published in the literature and are commercially available in software packages [143]. Unfortunately, the typical accuracy errors between these in silico LogP, LogD, and pKa models and experimental data are somewhat large ranging between 0.1 and 1 log unit. While this type of error does not interfere with prioritizing hits into a high/low classification scheme, caution should be taken by medicinal chemists when ranking hits in a narrow range (i.e., high to low) using calculated LogP, LogD, and pKa values. The prediction of aqueous solubility of a compound, from its structure, is a much more challenging problem than the prediction of lipophilicity since aqueous solubility involves the partitioning of the solid form of the compound into water [139, 144, 145]. The solubility of a compound depends on the charge state of the molecule, the pH of the solvent, its solid-state form (i.e., amorphous or crystalline) and the temperature of the surrounding environment. The prediction of the solid-state form of a compound from its structure has not been solved [139]. In addition, typical experimental errors for aqueous solubility measurements can be approximately ±1.0 LogS units or larger. Thus, calculating the contributions of structural units within a compound to predict quantitatively solubility has a great deal of error embedded in it. Calculated intrinsic solubility values of hits using empirical-based models should not be used at key decision points in the drug discovery process. The quantitative structure–property relationship (QSPR) and quantitative structure–activity relationship (QSAR) approaches are regression models that relate a set of compound predictors such as, physicochemical properties (i.e., LogP and MW) and theoretical molecular descriptors (i.e., molecular orbital energies) to the compound’s property of interest [146]. The QSAR approach has been used many times to predict intrinsic aqueous solubility (neutral form of the molecule) with models ranging from simple linear models using small sets of molecular descriptors to non-linear models using large sets of molecular descriptors [146-150]. It is generally concluded that these types of QSAR or QSPR methods cannot be used to globally predict aqueous solubility for pharmaceutical compounds with sufficient accuracy due to deficiencies in the algorithms and​/or descriptor sets of the individual models [147, 150]. Calculated intrinsic solubility values of hits using QSAR or QSPR models should be considered only a very rough estimation of their true value. An interesting approach for improving the prediction of aqueous solubility of virtual compounds is based on the idea of using several QSAR models in combination with a random forest classification approach [151]. In this machine learning application, ten commercially available QSAR models were used to generate aqueous solubility data for all compounds. The best QSAR model was selected for a virtual compound based on the predictions for structurally similar compounds using decision trees generated by the random forest method. In general, in silico solubility predictions cannot currently be used as a substitute for experimental measurements at key decision points in the drug discovery process.

In Silico ADME: A in ADME

In silico ADME has been reviewed many times by others and readers are encouraged to examine these publications [114-130]. The A in ADME represents the absorption process which involves the extent and the transfer rate of a compound from the gastrointestinal (GI) fluid across primarily the jejunum and the ileum segments of the small intestine into the portal blood system. The driving forces for this process are the compound concentration gradients between the GI and the portal blood supply, electrical differences, the hydrostatic pressure gradients introduced by the presence of the compound and external factors such as dissolution rates of the compounds, stomach emptying rates and food consumption. For a compound to reach the general systemic circulation, it must penetrate cell barriers either by passing between cells (paracellular route) or through them (transcellular route). When compounds utilize the transcellular route, they cross the membrane barrier either by mechanisms involving the active participation of components of the membranes (i.e., energy-consuming carrier-mediated processes) or by passive processes (i.e., non-energy-consuming diffusion processes). While carrier-mediated transmembrane processes play an important role for some compounds, passive diffusion through the bilayer membranes is the dominant process in the disposition of most small molecules.

The extent of absorption of a compound is measured as the percentage of orally dosed compound absorbed (%Fa) across the small intestine into the portal blood system. In a typical experiment, the concentration of the compound being orally administered is known. By removing a blood sample via a catheter inserted into the hepatic portal vein, the concentration of the compound in the portal blood system can be determined. The %Fa value is calculated as the ratio of the orally dosed concentration of compound divided by the portal vein concentration of compounds. %Fa can be considered a function of two key components: the permeability and the solubility of the compound. The permeability of a molecule is a measure of the ability of a molecule to cross a cell membrane barrier and is expressed in velocity units. Permeability is primarily a function of the compound’s LogD, MW, and the aqueous diffusion constant of the compound. Various in vitro permeability assays have been developed that mimic the relevant characteristics of in vivo absorption including the human colon adenocarcinoma (Caco-2) cell assay and the Madin-Darby canine kidney (MDCK) cell assay. These will be discussed along with other methods to determine permeability in a later section.

There is a great deal of interest in in silico methods for predicting the %Fa or the permeability of hits from their 2D chemical structures [152-156]. Many of these methods use QSAR models with molecular descriptors (i.e., polar surface area, molecular size, flexibility, hydrogen bonding capacity and so on) along with a variety of calculated physiochemical properties such as LogP and pKa [148]. In addition, these models use building methodologies such as linear or non-linear regression techniques, classification trees, partial least squares, and neural networks. Generally it is found that these QSAR %Fa or QSAR permeability models predict the training data set reasonably well, but typically do not predict new data sets with acceptable accuracy [157]. This result is not unexpected for %Fa since it is not a discrete fundamental property of the compound and, therefore, cannot be predicted solely from a structural representation of the compound. %Fa is a complex biological process highly dependent upon many internal and external factors. However, the permeability of a compound is a discrete property of the compound and accurate predictions of it by in silico methods may be developed in the future. In general, in silico absorption predictions cannot currently be used as a substitute for experimental measurements at key decision points in the drug discovery process.

In Silico ADME: D in ADME

The D in ADME represents the distribution process of a compound in the body. When a dose of a compound enters the systemic circulation (i.e., blood), the compound is distributed to all parts of the body. A theoretical volume of distribution at steady state (Vdss) is assumed as the volume into which a drug is distributed in the body. Thus, a high value of Vdss (> 42 L) for humans indicates that a compound is highly likely to be distributed throughout body tissues whereas a low value (< 3 L) suggests it is predominantly located in the systemic circulation. The Vdss in conjunction with the clearance of the compound determines the half-life of a compound in the body. In addition, the distribution of a compound is an important parameter to understand since it is generally a prerequisite for a compound to pass from the blood into other fluids and tissues before it can induce/exert its pharmacological action. The Vdss is a function of the tissue to plasma binding capacity of the compound as well as the ease with which the compound crosses the membrane and to the organ/tissue blood perfusion rates. The binding capacity of a compound is related to its physicochemical properties (pKa, Log P, aqueous solubility, etc.) and the affinity property of the compound for organs/tissues (blood protein binding and cell binding). In general, there is a large variability in the Vdss parameter in humans and animals. The clinical error range in Vdss human data is typically 2- to 3-fold from the mean values. There is a great deal of interest in predicting Vdss from chemical structures [158-165]. For example, Lombardo and co-workers [157, 161] have used a variety of QSAR linear and non-linear statistical techniques to predict Vdss from chemical structures. They have shown that Vdss can be predicted within a geometric mean 2-fold error. This magnitude of the error is on the same order as experimental data and can be used as a substitute for experimental measurements at key decision points in the drug discovery process.

In Silico ADME: M in ADME

The M in ADME represents the metabolism of a compound by enzymatic reactions. Hydrophobic compounds usually have low aqueous solubility under physiological conditions and must be converted into more hydrophilic molecules before they are excreted from the body. This biotransformation is typically denoted as Phase 1 and Phase 2 enzymatic reactions. The Phase 1 enzymatic reactions convert the hydrophobic parent compound to a more polar metabolite by oxidation, reduction, or hydrolysis chemical reactions. These reactions expose or introduce a polar functional group (-OH, -NH2, -SH, or -CO2H), and usually result in only a small increase in the hydrophilicity of the compound. During the last four decades, extensive studies of Phase 1 reaction types have shown that there are many cytochrome P450 enzymes (CYP450s) [166] and non-P450 enzymes (i.e., flavin monooxygenases (FMOs) and monoamine oxidases) [167-169] that have very broad substrate selectivity, catalytic versatility, and play important roles in the xenobiotic metabolism. The CYP450s are the predominant pathways for drug metabolism of most small molecule drugs, and the seven most common CYP450s are 1A2, 2B6, 2C8, 2C9, 2C19, 2D6, and 3A4. Phase 2 drug-metabolizing enzymes are generally transferases and are responsible for conjugating a compound or its Phase 1 metabolites with a highly polar molecule such as, glucuronidation, sulphation, and acetylation where they are mediated by the enzymes glucuronyltransferases (UGT), sulfotransferases (SULT), and N-acetyl-transferases (NAT), respectively [170]. The ability to predict the metabolic fate of a compound using in silico methods would allow medicinal chemists the ability to establish drug-design strategies that increased the half-life of a compound in the body or to eliminate reactive metabolites that might exhibit toxicity.

Several computational approaches have been created for predicting the metabolic fate of a compound. Expert systems are software packages that rely on a set of programmed rules, which have been distilled from available metabolism/ toxicological knowledge and human expert judgments to predict metabolites and their toxicity. This computational approach has been more successful than other approaches since it is intuitively appealing to most users. Software packages, such as TIMES, METEOR, METASITE, and METABOLEXPERT are commercially available [170-176]. These types of expert models typically predict a large ensemble of metabolites that are possible for a compound. In many cases, the ensemble of metabolites contains most of the