Sustainable Practices in Medicinal Chemistry Part 2: Green by Design

Apr 4, 2017 - Software from major molecular modeling software companies such as Biovia, BioSolveIT, Chemical Computing Group (CCG), OpenEye, and Schro...
1 downloads 17 Views 3MB Size
Perspective pubs.acs.org/jmc

Sustainable Practices in Medicinal Chemistry Part 2: Green by Design Miniperspective Ignacio Aliagas,† Raphael̈ le Berger,‡ Kristin Goldberg,∥ Rachel T. Nishimura,⊥ John Reilly,# Paul Richardson,∇ Daniel Richter,∇ Edward C. Sherer,§ Brian A. Sparling,○ and Marian C. Bryan*,† †

Genentech, Inc., 1 DNA Way, South San Francisco, California 94080, United States MRL, Merck & Co., Inc., 2015 Galloping Hill Road, Kenilworth, New Jersey 07033, United States § MRL, Merck & Co., Inc., P.O. Box 2000, Rahway, New Jersey 07065, United States ∥ Innovative Medicines Unit, AstraZeneca, Building 310, Milton Science Park, Cambridge, CB4 0FZ, U.K. ⊥ Janssen Research & Development, LLC, 3210 Merryfield Row, San Diego, California 92121, United States # Novartis Institutes for BioMedical Research, 250 Massachusetts Avenue, Cambridge, Massachusetts 02139, United States ∇ Pfizer Global Research and Development, 10777 Science Center Drive (CB2), San Diego, California 92121, United States ○ Amgen, Inc., 360 Binney Street, Cambridge, Massachusetts 02142, United States ‡

ABSTRACT: With the development of ever-expanding synthetic methodologies, a medicinal chemist’s toolkit continues to swell. However, with finite time and resources as well as a growing understanding of our field’s environment impact, it is critical to refine what can be made to what should be made. This review seeks to highlight multiple cheminformatic approaches in drug discovery that can influence and triage design and execution impacting the likelihood of rapidly generating high-value molecules in a more sustainable manner. This strategy gives chemists the tools to design and refine vast libraries, stress “druglikeness”, and rapidly identify SAR trends. Project success, i.e., identification of a clinical candidate, is then reached faster with fewer molecules with the farther-reaching ramification of using fewer resources and generating less waste, thereby helping “green” our field.



INTRODUCTION Development of new synthetic methodologies including both early and late stage diversification strategies has expanded the medicinal chemist tool box used to construct molecules, swelling the pool of potential molecules. With this, the environmental impact, which is the amount of chemicals, rare-earth metals, energy, and solvents used and the waste created, expands as well.1−3 Such impact should not be disregarded under the mistaken belief that only process chemistry1,4−6 and manufacturing matter but should hold medicinal chemistry to the principles of green chemistry as well if our field is to continue in a sustainable manner.6−8 While it is often assumed that the impact of medicinal chemistry is trivial compared to the impact of development and manufacturing activities, it has been estimated that drug discovery is responsible for 200 000 to 2 million kg of waste with a further 150 000 to 1.5 million kg during the preclinical process.9 Consequently, there is an increasing need to refine what can be made to what should be made while maintaining Pharma’s focus on patient health and innovation.10 The review presented herein summarizes multiple cheminformatics approaches in early drug discovery that together help triage designs, rapidly © 2017 American Chemical Society

identify key structure−activity relationships (SARs), and influence the likelihood that high value molecules are synthesized early, improving our efficiency in identifying a clinical candidate and lessening our environmental impact along the way. By working to reduce the overall number of compounds required for a given target/clinical project, this review seeks to build on the authors’ previous work within the Medicinal Chemistry Sub-Team of the ACS GCI Pharmaceutical Roundtable,11 focusing on fewer, more “high value” analogs leading to inherently more efficient and greener projects and a pharma industry as a whole. With increased awareness of the chemicals required including catalysts and solvents and the waste created, there has been a push within the industry to “green” our chemistry, namely, using more chemically efficient routes that produce less waste.11 An innately greener strategy, though, would be to take these more sustainable strategies and apply them to making only those molecules that would have the strongest impact on the project’s goal of delivering a candidate.11,12 Previously, the Received: December 14, 2016 Published: April 4, 2017 5955

DOI: 10.1021/acs.jmedchem.6b01837 J. Med. Chem. 2017, 60, 5955−5968

Journal of Medicinal Chemistry

Perspective

Figure 1. Application of computational chemistry in an efficient and sustainable or “green by design” manner with the goal of shortening the time and total number of compounds required to reach the clinical candidate. Starting with the large amount of possible compounds that could be made, a minimum number of more “drug-like” molecules are then prepared to test key hypotheses and filtered again to allow for candidate quality molecules to be generated in the least iteration, increasing speed and improving the “green” footprint of the project and chemistry as a whole.

standard paradigm for medicinal chemistry resembled that shown at the left of Figure 1 where a broad set of possible compounds were prepared over time, eventually leading to a clinical candidate. In this paradigm, there was no discussion of refining the possible compounds for “druglikeness”, i.e., metabolic stability, solubility, kinase selectivity, or the sustainability of the chemistry. Applying filters to refine the possible compounds into those with the highest probability of generating a clinical candidate shifts this strategy to the “green by design” paradigm shown on the right of Figure 1. Often, chemical matter is loosely categorized as either “hitlike” or “lead-like.” In both cases, the medicinal chemist has tools to efficiently progress chemical matter and identify a clinical candidate. For “hit-like” matter, ligand efficiency metrics are often used to help rationalize differences between molecules and series. In these situations, synthetic library techniques are often used to quickly build molecular complexity13 or interrogate a specific structural feature that may comprise a key component of the pharmacophore.14 When compounds display properties that are “closer” to the required characteristics of a lead molecule, additional computational chemistry techniques may be applied including filters for in vitro metabolic stability and physicochemical properties,15,16 molecular modeling/docking,17 CNS predictability,18 and off-target promiscuity prediction.19 As a project matures, the computational predictions may be compared to real data. This will either improve computational model accuracy or at the very least inform teams that certain models do not work well and should not be trusted in decisions regarding potential new compounds. A potential flow scheme for this “green by design” paradigm is shown in Figure 2. Beginning with virtual screening, pharma is seeking to refine their approaches from early stage hit finding to more focused, “lead-like” screening libraries20 and structurebased virtual screening with an eye on desirable physicochemical properties.21 Upon screening completion, hits may be triaged based on calculations such as ligand efficiency (LE) and lipophilic ligand efficiency (LLE) among other parameters like absorption, distribution, metabolism, and excretion (ADME) properties and solubility predictions. Directed library design, quantitative structure−activity relationship modeling (QSAR), and multiparameter optimization (MPO) then identify the most ideal starting points.22 The minimal number of molecules are then prepared and filtered again to allow for lead-like

Figure 2. Potential flow scheme for using computational chemistry and cheminformatics to “green” the standard medicinal chemistry paradigm. Applicable tools are shown at the right of each stage of the medicinal chemistry project, from hit finding to lead finding through candidate finding with the ability to reapply earlier filters as well.

molecules and then candidate-like molecules to be generated in the least iterations possible. Following synthesis and biological testing, data visualization and analysis tools allow the chemist to rapidly identify areas of the molecule that show interesting SAR such as activity cliffs23 for further development. These tools include data visualization software, QSAR, 24 and multiparameter optimization (MPO).25−27 Filters can be reapplied and new filters added to survey molecular complexity, investigate off-target prediction and synthetic strategy, and optimize CNS penetration or ADME properties as well as refine LE, LLE, and MPO. Finally, with the lead in hand, methods to evaluate synthetic efficiency such as retrosynthetic analysis and diversity-oriented synthesis (DOS) are being employed with consideration of sustainability,4−6,28,29 thus enabling the acceleration of follow-up molecule design as well as progression of candidates from medicinal chemistry into process chemistry and ultimately the clinic.30,31 By application of filters at each step prior to initiating synthesis, the number of compounds and time overall are diminished between inception and candidate selection, shortening the arrows at the right of Figure 1 and improving the “green” footprint of the project and chemistry as a whole. The combined application of the tools and methods mentioned above and highlighted below can affect the 5956

DOI: 10.1021/acs.jmedchem.6b01837 J. Med. Chem. 2017, 60, 5955−5968

Journal of Medicinal Chemistry

Perspective

Figure 3. Example of lipophilic ligand efficiency (LLE) map, generated by plotting pIC50 against the cLogP, for two chemical series where series A is shown in orange and series B is shown in blue and the lines corresponding to LLE integers are shown. In this case, series A is more favorable than series B based on improved LLE (LLE 6−8 versus LLE 2−7). The green arrow indicates a structural change that improved potency without affecting cLogP. The red arrow is an example of a structural change that results in a decrease in cLogP without affecting potency. The purple arrow shows improved potency with a marginal increase in lipophilicity.

Ligand Efficiency (LE) and Lipophilic Ligand Efficiency (LLE). From project inception, design metrics and tools have proven critical to aiding project teams in making smarter and more efficient decisions by providing a framework for evaluating diverse and oftentimes competing dynamics. In triaging a HTS campaign, there are multiple factors to consider when selecting scaffolds and compounds to pursue including physicochemical properties, desirable on-target potency, available SAR, and synthetic considerations. All can be critical components in ensuring that the discovery process begins in the most drug-like property space available. As such, efficiency indices have proven to be key tools in the medicinal chemistry tool box. When evaluating broad numbers of scaffolds from screening campaigns, efficiency indices provide a quantitative comparison beyond just potency against the target, in particular by adjusting for size and lipophilicity. These indices, when properly validated,34 can provide a useful triage strategy by distinguishing between an almost infinite combination of library molecules and potential data filters by highlighting and weighting on physicochemical properties. The influence of physicochemical properties such as molecular weight and lipophilicity on compound attrition is well-known with lower molecular weight and less lipophilic compounds corresponding to higher likelihood of progression into and through the clinic.35 Since hit-to-lead processes and lead optimization often increase potency through increasing both of these metrics, it is prudent to begin in property space that affords potential for growth. Although a large number of efficiency metrics exist in the literature, LE and LLE stand out as having been widely adopted in industry to effectively tackle this task. It is generally accepted in the field that higher LE and LLE values are desired with LE > 0.3 and LLE > 5 being ideal.20 LE is defined as the binding energy of a ligand per non-hydrogen atom for its substrate and is solved by multiplying 1.37 with

decision-making process such that a higher proportion of compounds with a desirable combination of target activity and drug-like properties are efficiently prepared in an environmentally sustainable manner. This in turn provides a higher probability of success in moving a compound forward to development while significantly reducing the number of compounds needed to achieve optimization objectives and/or candidate selection and in so doing decreasing the chemicals, energy, and resources consumed and the waste generated. As each of these cheminformatics approaches are applied, medicinal chemists operate in an increasingly greener and sustainable manner. Below, these methods are highlighted in order that they are commonly first used in this paradigm with the understanding that filters and visualization strategies need to be constantly refined and validated and are repeatedly applied throughout the life of a project.



STRATEGIES FOR HIT FINDING

Application of cheminformatics tools and the ability to analyze large quantities of data lead to shorter design cycles, faster optimization and can reduce the number of compounds needed to achieve optimization objectives and/or candidate selection, resulting in a greener chemistry campaign. In order to handle the complex problem of interpreting large data sets covering multidimensional parameter optimization such as those at the culmination of a high throughput screening (HTS) campaign, chemists benefit from cheminformatics and visualization tools, which help clarify analyses and strategies. Owing to the expansion of the field of cheminformatics,32,33 data analysis tools now include visualization, machine learning, and computational chemistry. Since it is not possible to provide an exhaustive review of specific tools, techniques, or processes, an overview of general concepts is provided below, starting with efficiency indices. 5957

DOI: 10.1021/acs.jmedchem.6b01837 J. Med. Chem. 2017, 60, 5955−5968

Journal of Medicinal Chemistry

Perspective

nonexperts to run preset calculations on demand. Making these accessible to nonexperts allows for greater time and resources to be devoted to continued model expansion and new model development. Software from major molecular modeling software companies such as Biovia, BioSolveIT, Chemical Computing Group (CCG), OpenEye, and Schrodinger offer the user the ability to load large data sets to enable molecular modeling, visualization, machine learning, and workflow informatics. Loading data into visualization tools presents the chemist with a highly customizable environment for building detailed visualization of trends in structure and activity. Increasingly, discovery projects will have data automatically updated in graphical representations to aid in the design of new molecules and synthetic target planning. A significant improvement in the interpretation of multiple end points is the application of MPO.25,27,43 In MPO, either intuitively defined or machine-learning-derived relationships are built from several molecular properties or assay end points to assess the desirability of individual compounds or chemical series, thus allowing for a balance of properties, not disproportionately placing weight toward one term such as on-target activity. By graphically viewing trends (such as trends in MPO score) across chemical series instead of one single database record at a time, chemists are able to visualize progression being made toward specific design goals. In addition, deconstruction algorithms in which fragment-derived chemical series are split into central scaffolds and peripheral substitutions allow chemists to automatically generate SAR analysis and visualizations. Tools can then provide a hierarchical tree of substructures from the molecules, which includes scaffolds as well as functional groups that contribute to the SAR. Matched molecular pair (MMP) analysis has proven to be a valuable aid in interpreting SAR44−47 as it mines large data sets for pairs of molecules that exhibit a single-point structural change and then selects for pairs for which that change results in a consistent alteration in properties and/or activities. By use of this methodology, trends for on- or off-target activity versus properties can be easily identified. For instance, it is expected that an MMP analysis on a large data set of human ether-a-gogo-related gene (hERG) off-target activity would show a significant decrease in hERG activity when nonpolar groups are replaced with more polar groups.48 Historically, some of these tools have been difficult to use, limited to expert users, and even then only sporadically provided useful output. Thus, these techniques have only slowly gained traction in medicinal chemistry groups. In the past decade, cheminformatics has helped to bridge many of these gaps and has provided tools that are easy to use, centrally located, and consistent across platforms. As these tools have been refined and optimized in recent years, they are becoming faster, more convenient to access, and more intuitive. This has led to increased uptake and consequently increased impact. With the ability to rapidly analyze the data, a scientist can now validate or disprove a hypothesis more rapidly without wasting additional resources. Directed Library Design. With the development of highthroughput chemistry in the 1990s, medicinal chemists broadly adopted parallel synthesis and library preparation as tools to generate large numbers of analogs for rapid SAR determination. At that time, the library size was often regarded as more important than the properties of the newly synthesized compounds, and monomer selection was based primarily on

pIC50 over the number of non-hydrogen atoms in the molecule where pIC50 is the −log10(IC50) (for derivation, see Hopkins et al).36 LLE, though, is the value measured by subtracting cLogP from pIC50. It is of note that the IC50 can be replaced with EC50, Ki, or Kd in these formulas. These two metrics can then be used in HTS triage to favor smaller and more hydrophilic hits. LE is particularly useful when analyzing hits from a fragment-based screen, where compound libraries have low cLogP and differences in potency yield larger and therefore more distinguishable changes in LE than they would for larger compounds.37 After hit deconvolution, the use of LLE can continue to guide strategy from lead series identification to candidate selection. It is a particularly meaningful assessment in the lead optimization phase because lipophilicity influences a number of a compound’s traits including off-target activity and potential for further development.38 Plotting the pIC50 against cLogP for a series of compounds provides an LLE map (Figure 3). The LLE map illustrates the absolute and relative qualities of series, subseries, or individual compounds, and the arrows highlight trends. The green arrow in Figure 3 shows a promising new subseries to investigate, while the red arrow is an example of a structural change that reveals a region of the molecule where polarity is tolerated. The purple arrow illustrates that a minimal increase in lipophilicity may be justified given the substantial increase in observed potency. The use of LE and LLE provides practical guides for judicious selection of medicinal chemistry starting points and insightful monitoring of SAR in the series and candidate selection phases of drug discovery, and each has been successfully applied in several reported cases.39−42 Despite the simplicity of efficiency metrics, there is good correlation between a compound’s LE or LLE index and its probability of success in launching to market,36 highlighting the power of these metrics to enable expedited drug discovery. Data Analysis Tools: QSAR, MPO. With corporate databases in place, graphical interfaces can enable medicinal chemists to mine data using structural queries. One simple method to visualize data is to pair a structure to all known chemical and biological data for a given compound. This allows users to scrutinize database records one compound at a time. While this is limited in scope to a single compound, the depth of data that are accessible is usually quite valuable in identifying aspects that require improvement. It is equally necessary to have tools that can query multiple compounds. In these applications, instead of bringing up one structure and all its associated data, the user specifies a smaller number of records and a correspondingly smaller quantity of data for retrieval. One example of such a process is to pull all compounds registered for project X in the past 4 months and simultaneously retrieve all of their data in key assays. There exists a number of relevant software for this level of complex data analysis with some highlighted to show their scope. However, this is by no means a complete list. Indeed, this field is constantly expanding, and novel and impactful software will make some methods out-of-date for any manuscript by publication. These interfaces generally allow the export of large subsets of compounds to allow for tasks such as QSAR model building, property calculations, or 3D molecular modeling. While there are multiple, popular interfaces for performing descriptor-based QSAR studies (e.g., Microsoft Excel, JMP statistical software), companies are increasingly building Web interfaces that allow 5958

DOI: 10.1021/acs.jmedchem.6b01837 J. Med. Chem. 2017, 60, 5955−5968

Journal of Medicinal Chemistry

Perspective

Figure 4. (A) Distribution of measured stability, represented by human liver microsome predicted hepatic clearance (HLM) and colored according to stability (stable or less than 30% liver blood flow, green; moderate or between 30% and 70% liver blood flow, yellow; labile or >70% liver blood flow, red) by registration year. (B) Distribution of measured kinetic solubility of compounds by registration year and colored by kinetic solubility category (red, 40 μM).

commercial availability. However, in 1996 the Lipinski “rule of 5” refocused the parallel synthesis community toward the design of compounds with more desirable properties. The integration of in silico tools and library design has had a dramatic impact on library preparation.49,50 Today, large virtual libraries are created and computationally evaluated in silico, and a reduced number of compounds with good predicted properties are physically prepared and tested.51 This lessens the environmental impact of library impact by eliminating generation of compounds with low likelihood of success (and their resultant waste). While such strategies have drawn both praise and criticism concerning their reliability, limitations, successes, and failures, the implementation of best practices has been reported aiding in their fully informed and objective implementation.24 A number of commercial software packages are available for QSAR, visualization, and MPO calculations, but only a few have all these functionalities integrated into one interface devoted to library design (e.g., StarDrop by Optibrium, MOE by Chemical Computing Group, Maestro by Schrödinger, SYBYL by Tripos, and ROTATE by Molecular Networks GmbH).52 Many pharmaceutical companies have opted to develop their own customized in-house versions as well. As a general approach to in silico library design, the process starts with a generic chemical transformation of interest, showing the reaction inputs and the resulting products. The second step is enumeration where the software, which is linked to a database of commercially available monomers, can enumerate the reaction to generate a virtual library. Different filters can be applied at this stage. For example, monomers that contain multiple functional groups which may be incompatible with the reaction conditions or known toxicophores can be removed. Monomers can also be selected based on known reactivity in such transformations, as in the Pfizer global virtual library.53 Once scaffolds have been identified from a HTS campaign, virtual libraries can be designed and enumerated and in silico predictions can be performed to assess the physicochemical and ADME properties of potential molecules for each scaffold. Constraints can be applied to the enumerated products, and only compounds that fall within the desired parameters will be made and assayed. For more mature series, virtual libraries can be built and the products run through established program QSAR, docking models, or Free-Wilson analysis. By enabling the virtual screening of thousands of

compounds and the rapid identification of ones with favorable predicted properties, directed library design reduces the number of compounds requiring synthesis, increasing project efficiency while decreasing consumables and waste, aiding overall sustainability even from the earliest stages of a project.



STRATEGIES FOR LEAD FINDING

ADME and Solubility Predictions. With the data analysis tools in hand, projects can use models and generate filters to further improve scaffolds and determine the likelihood of success without having to synthesize multiple generations of analogs, saving time and resources. Two such interwoven areas where the above-mentioned data analysis tools have been critical are in the areas of in vivo stability and solubility. For example, kinetic solubility and liver microsomal clearance can be measured for all compounds registered. These so-called “first tier assays” generate very useful trends and are used as broad filters before compounds progress further in the drug discovery paradigm, limiting potential scale-up waste for compounds that show low likelihood of desirable exposure. Computational models are available for both of these parameters prior to initial synthesis during the design phase and have had a marked impact on the molecules that are generated at Genentech and in the industry as a whole (Figure 4).16,17,54−57 The yearly profile of predicted hepatic clearance in human liver microsomes (HLM) improved significantly after the implementation of a HLM QSAR model at Genentech that has been used as a synthetic filter since 2010, more than doubling the proportion of low clearance (stable) compounds from previous years (Figure 4A, 4% stable in 2009 versus 20% stable in 2010). While the percent of stable compounds has not continued to increase, this is to be expected as new scaffolds are discovered, the model is refined, and molecules and libraries are designed with a focus of testing a key hypothesis rather than focusing solely on metabolic stability. In light of this, it is important to note that the percent of stable compounds has not receded either. Like the HLM calculator discussed above, a solubility index (SI)56 score has been routinely used at Genentech since 2011, the effect of which can be seen in the improved kinetic solubility after 2011 in Figure 4B. Similarly, a kinetic solubility QSAR model was introduced in 2015 as a synthetic filter which has resulted in an additional increase in more soluble compounds from previous years (Figure 4B). In addition, 5959

DOI: 10.1021/acs.jmedchem.6b01837 J. Med. Chem. 2017, 60, 5955−5968

Journal of Medicinal Chemistry

Perspective

using machine learning methods. An example of this would be the Merck Research Laboratories (MRL) system of electronic counterscreening (eCS),19 an alternative to using external vendors to screen all compounds (see section on “Predicting Off-Target Screening Activities”). Application of molecular modeling can also help chemists optimize away from AMES liability.62 Together, these two examples highlight how data analysis tools can lead to more drug-like chemical matter. The application of such predictors and filters then allows project teams to access candidate-like molecules in fewer iterations, leading to a greening of projects and programs as a whole. Diversity-Oriented Synthesis. As a continuation of the directed library design discussed above, diversity-oriented synthesis (DOS) is another strategy that seeks to generate libraries of structurally complex and functionally diverse small molecules in an efficient manner. In much the same way as the standard paradigm describe in Figure 1, traditional library synthesis focused more on size with large numbers of structurally similar compounds. The paradigm is now shifting toward particular emphasis on increasing a library’s structural and functional diversity efficiently. DOS has been defined as deliberate, simultaneous, and efficient synthesis of libraries in a diversity-driven approach. The libraries yielded are normally smaller in size than their commercially available libraries but are typically structurally more complex, with a greater variety of core scaffolds and richer stereochemical variation.63 Diversityoriented synthesis (DOS) generates small molecule libraries with high levels of structural diversity with examples highlighting green chemistry in the literature.28,63−66 DOS can take into account both standard drug-like chemical space and “natural-product”-like space with most DOS library design strategies leveraging information about existing biologically active small molecules to generate compounds that similarly target these regions.67 Libraries can be filtered as described above for desirable physicochemical properties. Because DOS libraries survey broader structural diversity, they are already impacting “undruggable” or more difficult targets such as recent reports around mitosis modulators, histone deacetylase inhibitors,68antimalarials, and treatments for Chagas disease.64,69−71 As such, DOS is an ideal combination of innovation and efficiency in both library size and chemistry, leading to fewer, more high-value compounds being generated with less waste. Structure-Based Design. Hand-in-hand with property calculators in this green paradigm is molecular modeling, which is a valuable component in the collaborative design of drugs. A full review on its application to drug discovery, optimization, and development is outside the scope of this work (for examples, see referenced literature).72,73 Instead, we seek to highlight the major objective for application of modeling services: to answer hypotheses, reach program objectives, and help explain structure, function, and reactivity in the most rapid manner with the fewest number of compounds needed. This reduction in compound number has a direct impact on the chemical footprint for a project, so modeling efforts in combination with the other strategies listed are at the forefront of green chemistry initiatives. In early stage discovery work, structure-based design is often a key focus of modeling requests. The prediction of drug− target binding can be accomplished through many techniques including docking and scoring, molecular mechanics interaction energies, hybrid quantum mechanics/molecular mechanics calculations, free energy perturbations (FEP), or machine

calculators for subsequent DMPK measurements (permeability, hepatocyte stability, plasma protein binding, etc.) are also available. This is not to say that the models provide a hard filter; application of computational models can differ depending on a project’s stage, the nature of the therapeutic target, and specific issues associated with each chemical series. The effect of these models on compound design and synthesis is best described within the framework of specific projects. The Janus kinase 1 (JAK1) inhibitor program at Genentech is one example of a project that successfully improved compound stability by using computational models. For JAK1, calculated HLM stability probability was used to prioritize compound synthesis in multiple chemical series, leading the team to efficiently use resources to support those series with higher predicted stability in vivo.58,59 In the early stages of optimization, the team’s primary focus was to establish SAR related to JAK1 inhibition, not to tackle their metabolic stability, which is illustrated in the poor measured stability seen at the left in Figure 5. As the team began to achieve consistent

Figure 5. Predicted probability of stability in the human liver microsomes assay for chemical series 2. The pie charts show the relative number of compounds, as the area of each pie slice, in the three measured liver microsomal categories (stable, S; moderate, M; labile, L) and binned by calculated probabilities (X-axis) for the two chemical series shown for the JAK1 project. The size of each pie is proportional to the relative number of compounds in each probability bin.

JAK1 potency, the use of the HLM stability QSAR model was used to screen synthetic targets, resulting in a larger proportion of analogs with high probability of microsomal stability. This translated to increasing numbers of experimentally stable compounds. The strong correlation between predicted probability of being stable and experimental HLM stability allowed the JAK1 team to prioritize compounds such that a large proportion of synthesized compounds were indeed experimentally stable. During the exploration of a new series for the NaV1.7 inhibitor program, Amgen utilized solubility predictors as integral tools for lead generation.60,61 A new series afforded highly potent hits that concomitantly suffered from limited solubility. Three positions amenable to analog diversification were assessed with the goal of improving physicochemical properties while maintaining a high level of potency against NaV1.7. In order to expedite this process, potential targets were weighed utilizing predictors for stability, described in the example above, as well as solubility in pH 7.4 PBS. Further, the in silico models were continually refined using newly obtained data. Another method of addressing ADME concerns is through the application of 2D or 3D molecular modeling including molecular docking, pharmacophore analysis, or machine learning/QSAR methods. Data sets representing individual chemical series, multiple project specific series, or global data sets covering entire corporate databases can all be modeled 5960

DOI: 10.1021/acs.jmedchem.6b01837 J. Med. Chem. 2017, 60, 5955−5968

Journal of Medicinal Chemistry

Perspective

Figure 6. Schematic representation of a conformer search used to determine a quantum mechanics (QM) derived Boltzmann population of conformers for molecular modeling applications. In this example the resultant population is used to calculate the vibrational circular dichroism spectra (VCD), shown at the bottom left.

molecular properties, selectivity, synthetic tractability, or intellectual property. Deconstructing hit compounds with information gained from fragment-based approaches or scaffold hopping may enable a team to more efficiently sample chemical space and avoid synthesizing large numbers of compounds. Moving past the macromolecular interactions, conformational analysis using molecular mechanics or quantum mechanics can provide useful information on structure, conformational flexibility, reactivity, and many other molecular properties. Reactivity predictions, which predict which site on a molecule will preferentially be susceptible for a given type of transformation, can greatly reduce the synthetic cost of performing tasks such as explicit metabolite identification and may benefit late stage functionalization, as reviewed by Cernak et al.,77 where a precursor compound can be modified at an advanced synthetic stage, saving chemical resources and cycle time. A discussion of such methods is outside the scope of the current review and will be covered in future work. Once candidate selection has been achieved, compounds commonly move into a development or process chemistry department where understanding chemical reaction mechanisms can help to optimize yield and minimize cost/ environmental impact with calculations such as process mass intensity (PMI).4,5 Molecular modeling may be applied to help rationalize reaction mechanisms, understand kinetic results, or contribute to catalyst or reactant/intermediate design. Important factors governing regioselectivity, reactivity, or enantioselectivity can be quantified and acted upon using quantum mechanics. Throughout the process of drug design and development, analytical chemists are highly engaged in compound characterization and benefit from collaborative application of modeling

learning methods. Each of these tasks can be enabled with crystal structures of drug−target complexes or by building homology models. Both allow docking and scoring to benefit from known or deduced interactions and are especially important when chemical series are highly divergent. Other techniques that aid in the identification of target binding and interactions may involve conventional biochemical and cell based assays, NMR, mass spectrometry, footprint analysis, surface plasmon resonance (SPR), or cryo-EM. AstraZeneca currently uses solution-state NMR analysis of high affinity compounds to complement homology models as a tool to aid design.74 Compound affinity can be increased through stabilizing the molecule in its biologically active conformation. This can be done, for example, through added rotational restriction by increased steric bulk75 or use of macrocycles to lock a compound in the desired conformation.76 This benefits synthetic chemists by providing rapid insight into the effect a structural change has on a molecule’s conformation and ultimately its potency and physical properties, leading to quick turnover of design ideas. Strong collaborative ties between synthetic chemists, structural scientists, and molecular modelers are paramount to program success and resource conservation. While the bulk of modeling and informatics work has been applied in the earlier stages of this workflow (see Figure 2), modeling techniques are being expanded to include lead optimization tasks as well. One important application utilizing the techniques described above is scaffold hopping. Fragmentbased screening and crystallography allow chemists to deconvolute what functionality is most required for activity. This technique allows for replacement of core scaffold structures with other species which might address issues of 5961

DOI: 10.1021/acs.jmedchem.6b01837 J. Med. Chem. 2017, 60, 5955−5968

Journal of Medicinal Chemistry

Perspective

tools such as quantum mechanics/density functional theory (QM/DFT) or QSAR/quantitative structure−activity relationship (QSPR). Spectroscopic analytical methods such as infrared spectroscopy (IR), electronic circular dichroism spectroscopy (ECD), NMR, and vibrational circular dichroism (VCD) are frequently compared to density-functional-theory-predicted calculated spectra. Application of VCD analysis allows chemists to accurately assign absolute configurations of compounds without small molecule crystals by comparing experimentally determined and calculated VCD/IR spectra. At MRL, the absolute configuration of more than 350 compounds has been assigned using VCD, representing a significant savings in smallmolecule crystallographic screening and effort needed to obtain single crystal structures (Figure 6).78 By establishment of the absolute stereochemistry, SAR trends can be better understood and synthesis directed toward compounds with higher likelihood of success, thus saving resources. Predicting Off-Target Screening Activities. Managing off-target activity is a crucial part of any medicinal chemistry program. In order to assess off-target liability early on, compounds are often submitted to a counterscreen panel to determine potential toxicity (e.g., ion channels for cardiac toxicity) or the potential to induce drug metabolism changes and drug−drug interactions (e.g., cytochrome P450 induction/ inhibition). All compounds of interest can be submitted to these panels, but it is a costly and wasteful process since the majority of the compounds tested turn out to be inactive in these off-target assays. The eCS tool leverages QSAR models to reduce the number of assays run by making a recommendation as to whether a given compound should be tested against a certain off-target.19 Only compounds with predicted risk of offtarget activity or uncertain predictions are selected for testing in the particular off-target assay. This tool not only utilizes QSAR models to predict the activity of a compound against a given off-target but also assesses the reliability of the prediction by analyzing historical activity prediction versus experimental data for the project of interest. These analyses can be visualized in a risk−benefit plot. Users can select a cutoff for activity as well as the level of risk they are willing to take if a compound with offtarget activity goes untested. eCS is both a green and costsaving tool for reducing in vitro assay requests. This has led to quantifiable cost savings and represents a green solution to the otherwise blanket submission of all compounds. It is useful to perform occasional analysis of compounds eCS does not suggest in order to ensure model performance, though.

Figure 7. Which is more complex? Examples of compounds with the same molecular weight but vastly different molecular complexity in terms of chirality, functional groups, and facile synthesis.

target on complexity is difficult to model or predict, though efforts at Bristol-Myers Squibb have recently been made to do just that.80 The result of considering both the absolute (or intrinsic) complexity of a molecule and the synthetic challenges was labeled “current complexity” by Li and Eastgate.80 Their model involved quantification of the ease of synthesis and weighted terms intuitively correlated to molecular complexity such as number of chiral centers and atom counts. This approach builds off of earlier contributions that calculated complexity from weighted descriptors of the 2D chemical structure.29,81−90 While correlation of descriptors of molecular structure makes intuitive sense, it can be difficult to determine what weights to assign various descriptors. Knowing that synthetic advances will reduce synthetic complexity over time, some researchers have sought to focus on absolute complexity which can lead to alternative modeling approaches. It is clear that multiple chemists discussing the absolute complexity of a compound will have varied opinions owing to experience, synthetic training, and aesthetics. Inheriting an idea from Johnson & Johnson91 in which the expansion of the compound collection was enabled through a crowdsourcing exercise across chemistry, MRL pursued a largescale crowdsourcing exercise across global chemistry to help create a consensus ranking of absolute complexity.13 What Sheridan et al. concluded from that study is that complexity, when determined through a consensus ranking, can converge to agreed-upon quantitative values.13 When a cheminformatics tool to calculate an acceptable quantitative measure of compound complexity exists, many possibilities for data analysis become apparent. Large-scale calculations of complexity may be used to gauge differences in compound collections or patent estates. Molecular property filters can be employed to analyze high-throughput screening data sets where chemists look to identify which chemical series should move into lead optimization. Adding a molecular complexity ranking to this stage of the compound prioritization may help in identifying more facile chemistry. MRL has found it informative to track compound complexity for medicinal chemistry programs over time, noting that significant differences in complexity for different chemical series are evident. An example of a representative kinase program is shown in Figure 8. Since more complex molecules are, at least initially, more difficult to synthesize, having a means to rank compound classes or development compounds in an unbiased way allows chemists to make strategic decisions related to chemistry direction. More complex chemistry may slow down the



STRATEGIES FOR CANDIDATE FINDING Molecular Complexity. Working on more complex chemistries will at least initially require a larger investment of resources. While this may seem counter to the push toward sustainable chemistry practices, design efforts to improve synthetic routes during compound development can lead to environmentally friendly or at least benign chemistry which is more possible when the overall chemistry is less complex. Different chemists assign different meanings to the concept of molecular complexity. Compounds with the same molecular weight, number of atoms, and type of atoms can have very different complexities (Figure 7), depending on whether one is considering the complexity intrinsic to the chemical structure or how complex it is to synthesize that structure. When a compound is difficult to make, its synthesis is deemed complex. Over time, if advances are made to simplify that synthesis, the compound’s synthesis will become less complex.79 A moving 5962

DOI: 10.1021/acs.jmedchem.6b01837 J. Med. Chem. 2017, 60, 5955−5968

Journal of Medicinal Chemistry

Perspective

The initial synthetic challenge projects face is the fact that there are often multiple diverse chemical series under consideration identified as described above (HTS, file-mining, etc.). While this can be highly positive for an early stage project, efficient prioritization of how to prosecute the discrete series is required. Teams often carry out this process through an analysis of potency, selectivity, intellectual property considerations, and physical properties with little regard to synthetic viability. Omitting this can have implications not only on rapid advancement (or de/reprioritization) of series but also on a projects’ total synthetic resource. Early stage chemistry, during hit-to-lead or lead finding, often contends with lower yields, column chromatography, and chiral resolution, focusing on generating potential targets to determine their project worth and rarely considering environmental impact or sustainability. Teams will often select a series and carry out retrosynthetic analyses on what are perceived to be the targets of highest value (Figure 9, top). We would argue that rather than focus on a specific target, the questions should revolve around the series as a whole, providing a more holistic approach involving proactive synthetic analysis of the targets/ series being considered (Figure 9, bottom). Similar to DOS mentioned previously,28 this exercise should in many ways parallel the thought process of the design team in considering not only the immediate targets but also the future plans if the initially synthesized targets prove to be promising or successful. While proactive synthetic analysis may appear both inflexible and potentially inhibiting synthetic creativity, this is definitely not the case. A better way to view the approach is as planning syntheses with the expectation of variability as the work progresses. In fact, one could also argue that a retrosynthetic approach is inflexible since the end goal is the completion of a single molecule with scant consideration for the synthetic parameters required (e.g., steps, yield, reagents, waste). Proactive synthetic analysis provides a method for looking holistically at the series, evaluating the synthetic parameters mentioned above, considering the environmental impact of items such as solvents and purification while looking at alternatives such as telescoping chemistry1,3,11 and planning for success. While such strategies may seem counter to the hit-tolead or even lead-finding stage, there is value in considering this approach as early as scaffolds of interest are identified. In this way, key late-stage intermediates that are not progressed due to a number of project factors should be readily amenable to other projects. Translating this to a specific example, pyrimidines (Figure 10) have been widely utilized in numerous medicinal chemistry programs. Target molecule 1, an early lead within a drug discovery program, was initially accessed through three steps from the reported acid chloride, 4, which itself was made from the exhaustive chlorination of orotic acid. On paper, this route looks reasonable, and the ready availability of the precursor to 4 certainly presents potential for the program. However, closer evaluation of this starting material combined with the experience gained from the preparation of multiple analogues utilizing this chemistry revealed problems inherent with this route, raising questions over its long-term sustainability for the program. With a proactive approach, the project team decided to rework the synthesis with the goal of addressing the above problems while also identifying a synthetic sequence fit for analog generation, library production, and scale-up. Starting from the readily available sulfone, 7, mild conditions were

Figure 8. Compound complexity for a discovery chemistry kinase program over time.

progress of optimizing SAR. Once compounds move into postcandidate nomination or development, there is ample evidence that more complex compounds have higher PMI such that ranking candidate compounds on complexity can inform on which programs may be more costly to pursue or have a larger chemical footprint when considering green chemistry initiatives.5,13 Retrosynthetic Analysis and Proactive Synthesis. The synthesis strategy in a discovery program will typically depend on the stage of the project. Preferably though, a program team would design the minimal number of compounds, which could be made by the simplest common route to progress a program into development. In reality, the complexity of the challenges involved coupled with iterative logistical design approaches dictates that this is never the case, and this uncertainty leads to an evolution of synthetic strategy as a program progresses. With this in mind, there are two critical synthetic assessments that can benefit a program from a green sustainability standpoint: • Minimum number of compounds of required molecular complexity to answer each specific design hypothesis; • Most efficient chemistry to disconnect a molecule in an orthogonal manner to enable rapid modification from an advanced intermediate (Figure 9).

Figure 9. Diagram comparison of synthetic planning paradigms. 5963

DOI: 10.1021/acs.jmedchem.6b01837 J. Med. Chem. 2017, 60, 5955−5968

Journal of Medicinal Chemistry

Perspective

Figure 10. Retrosynthetic approaches to substituted pyrimidines illustrating the potential benefits of a proactive synthetic approach.

numbers of reactions/operations predevelopment and the production of less waste. Inevitably, there will be some changes required in order to streamline productivity, yet this can lead to significantly accelerated timelines for a program.

identified to enable alcohol displacement to access the ether, 6, thus avoiding epimerization. At this juncture, displacement with the piperidine occurs on the symmetrical 4,6-dichloro compound to access 5 with no potential for the formation of regioisomers. Finally, 5 could be converted to the desired target compound 1 through a palladium-mediated carboamidation. This proactive approach provided tangible results for the project as well as increased sustainability: a 2× increase in speed for analog output over a 6-month period, the application of the methodology to two directed libraries, as well as the application on kilogram scale (37% overall yield) through mild chemistry and without chromatography. A further consideration once the core has been established is the inclusion of diversity vectors associated with the core. Keen awareness of disconnections at this juncture is important, which may enable variation as late as possible in the synthetic sequence. This enables the team to potentially stockpile a stable late-stage intermediate, which not only will allow for rapid versatile evaluation of a series but also will potentially expedite scale-up campaigns to compounds of interest. Although one can argue that identifying and making such an intermediate might require additional time for planning and enablement of the chemistry, the counter to this is that this time will be more than reclaimed through the more efficient SAR investigations and scale-up that follows. In addition, this approach can significantly set the chemistry for the series to a one-approach-fits-all, with the stage set for analog generation, library chemistry, and scale-up campaigns. This presents the opportunity to engage process chemists, who are already trained to consider the environmental sustainability of a route,4 at an earlier stage with the opportunity to streamline both the chemistry to access the core as well as the critical functionalization steps to introduce the various vectors with particular focus on isolations, order of the functionalization steps, process safety, and the availability of critical starting materials. This also allows for consideration of the environmental biodegradability of the core and final products.92 At many companies, early engagement with process colleagues is common practice such that examination of the synthetic process in this manner will lead to significantly reduced



CONCLUSIONS As the medicinal chemists’ tool box continues to grow, allowing for more and more complex molecules to be generated, the environmental impact expands as well. Furthermore, time and resources are finite and an increasing need exists to refine what could be made while focusing on innovation. As such, it is critical that there must also be an understanding of why we are making these molecules. The multiple cheminformatic approaches in drug discovery presented here seek to influence both design and execution and, in so doing, influence the likelihood that only high-value molecules are synthesized in the most efficient manner. This strategy of using computational methods to both design and refine molecules allows chemists to triage vast libraries early in a project, emphasize how likely a design will provide a “drug-like” molecule based on its calculated properties, and also highlight trends in the data to show when series or targets are more or less justified. While these methods must be viewed objectively in terms of their reliability and limitations, these methods provide tools for reducing the number of compounds required for the success of a project. An often overlooked and farther-reaching ramification of the widespread implementation of these tools, though, is that they allow medicinal chemists to become more efficient and “green” by omitting unnecessary molecules that waste precious resources. This is not to say that the chemistry being performed is inherently green but rather that we are limiting the burden of our field on the environment, allowing medicinal chemists to operate in an increasingly greener and sustainable manner.



AUTHOR INFORMATION

Corresponding Author

*Phone: +1 (650) 225-6532. E-mail: [email protected]. ORCID

Edward C. Sherer: 0000-0001-8178-9186 5964

DOI: 10.1021/acs.jmedchem.6b01837 J. Med. Chem. 2017, 60, 5955−5968

Journal of Medicinal Chemistry

Perspective

Marian C. Bryan: 0000-0002-3138-6888

with Lexicon Pharmaceuticals as Director of Process Chemistry for 8 years. Paul received his Ph.D. in Organic Chemistry from Sheffield University, U.K., where he worked with Professor Istvan Marko. He completed postdoctoral research studies at Exeter University, U.K., with Professor Stan Roberts and at the Scripps Research Institute, CA, under the direction of Professor Barry Sharpless. He is a member of the American Chemical Society and is an inventor or author of more than 60 patents and papers.

Notes

The authors declare no competing financial interest. Biographies Ignacio Aliagas earned his Bachelor’s degree in Chemistry from University of Valladolid (Spain) and his Master’s degree in Chemistry at San Francisco State University under Professor Scott Gronert. He joined Genentech as a computational chemist where he supported modeling efforts for the design of small molecule and peptide drugs using molecular dynamics, MM-PBSA, and free energy perturbation methods. He pioneered efforts at Genentech in drug properties prediction and DMPK end points through models for in vitro clearance, permeability, efflux, plasma protein binding, CYP time dependent inhibition, and log D. In addition he developed the framework for chemists to readily access calculated physicochemical properties, model prediction and validation. In addition, he supports therapeutic projects by providing modeling support in the optimization of drug design.

Daniel Richter is a Principal Scientist at Pfizer with 17 years of experience in Medicinal Chemistry. He earned a Master’s Degree in Organic Chemistry in 1999 from Illinois State University. He then began working as a medicinal chemist at Pfizer in Groton, CT, where he made significant contributions to the advancement of many oncology clinical candidates. He has been a green chemistry leader at Pfizer since 2004, and in 2006, he transferred to Pfizer, La Jolla, CA, and began leading the Green Chemistry team there from 2009 to 2012. He initiated the Medicinal Chemistry Sub-Team of the ACS GCI Pharmaceutical Roundtable in 2011 and continues to lead the team in promoting the implementation of green chemistry within the medicinal chemistry realm.

Raphaëlle Berger is a Senior Scientist in Medicinal Chemistry at MRL. In 2007, Raphaëlle received a Diplôme d’Ingénieur from CPE Lyon and a Master’s degree from the Université Claude Bernard Lyon 1 (France). She obtained a Ph.D. in Organic Chemistry from the University of Glasgow (U.K.) working under the supervision of Prof. J. Stephen Clark in 2011 and then undertook a postdoctoral fellowship at Princeton University with Prof. David W. C. MacMillan. She joined MRL in Rahway, NJ, in 2013 where she works on lead optimization programs.

Edward C. Sherer received his B.A. in Chemistry in 1997 from Lake Forest College. He pursued a M.Phil. at the University of Nottingham, U.K., under a Fulbright Fellowship from 1997 to 1998. He then joined the Cramer Group at the University of Minnesota where he continued research in quantum mechanics and molecular dynamics. After obtaining a Ph.D. in 2001 he joined Rib-X Pharmaceuticals in the search for novel antibiotics targeting the ribosome. In 2007 he joined MRL in Rahway, NJ, supporting discovery projects. Since 2011, he has directed the expansion of computational support of process chemistry at MRL. In 2016 he served as Chair of the COMP Division of the ACS after 8 years of service to the Division.

Kristin Goldberg is a Senior Scientist in Medicinal Chemistry at AstraZeneca. She received a Master’s degree in Organic Chemistry in 2006 from The University of Texas at Austin, working on natural product synthesis under Prof. P. Magnus. She moved to the U.K. in 2007 where she started her medicinal chemistry career at AstraZeneca working on diabetes medicines. She joined Cancer Research UK’s Drug Discovery Unit in 2012 where she worked on small molecule treatments for lung cancer for 2 years before rejoining AstraZeneca. She currently works in AZ’s Oncology iMED on early stage projects.

Brian A. Sparling earned his B.S. in Chemistry at Massachusetts Institute of Technology in 2008 prior to receiving his Ph.D. in Chemistry from Harvard University in 2013 under the direction of Prof. Matt Shair, where he accomplished an enantioselective total synthesis of hyperforin amenable to analog synthesis. Since then, he has been a member of the Department of Medicinal Chemistry at Amgen in Cambridge, MA, where his work has focused on the advancement of neuroscience programs. Since 2015, he has led Amgen’s medicinal chemistry Green Chemistry Team and is a member of the ACS GCI Medicinal Chemistry Roundtable.

Rachel T. Nishimura is a medicinal chemist at Janssen Research & Development, LLC in La Jolla, CA. She received a B.S. in Chemistry from Harvey Mudd College in 2009, where she worked on the green and biomimetic synthesis of natural products. Rachel was a chemist at the Synthetic Natural Gas Plant of Hawaii Gas before joining Janssen in 2010. Her research is focused on small-molecule programs in immunology, and she is a co-inventor on 13 patent applications. She has been a member of the ACS GCI Pharmaceutical Roundtable since 2014.

Marian C. Bryan earned her Bachelors of Science in Chemistry and Biochemistry from Clemson University prior to pursuing her Ph.D. in Chemistry under the direction of Professor Chi-Huey Wong at The Scripps Research Institute. Following an American Cancer Society postdoctoral fellowship with Professor Linda Hsieh-Wilson at The California Institute of Technology, she joined the Department of Medicinal Chemistry at Amgen, Inc., in 2006. In 2012, she joined Genentech, Inc., where she leads a chemistry team in Discovery Small Molecule Research.

John Reilly is a Senior Research Investigator in the Department of Global Discovery Chemistry at Novartis Institute of Biomedical Sciences in Cambridge, MA, U.S., where he leads the Chemistry Separations Group supporting Chemistry. His research goals include the promotion of biomimetic chromatographic affinity screens to investigate druglike properties of molecules, and he has published >25 articles in many analytical journals such as J. Chromatogr. A, J. Pharm. Biomed. Anal., Chirality, and Mol. Pharmacol. John obtained his MSc. and Ph.D. degrees within Analytical Science Imperial College (U.K.) in 2002 and has been a board member of the Chromatography Society in the U.K. and on the editorial board of “Chromatography Today” and “American Journal of Modern Chromatography.”



ACKNOWLEDGMENTS We thank the entire Medicinal Chemistry Sub-Team of the ACS GCI Pharmaceutical Roundtable, Kevin Freeman-Cook (Pfizer), Kate Holloway (MRL), Bob Sheridan (MRL), and Elisia Villemure (Genentech) for helpful comments.



ABBREVIATIONS USED DOS, diversity-oriented synthesis; HLM, human liver microsome; JAK1, Janus kinase 1; MMP, matched molecular pair;

Paul Richardson is the Director of Process and Analytical Technologies at Pfizer, La Jolla, CA. Paul has over 20 years’ experience in the pharmaceutical industry. He joined Pfizer in 2004, after working 5965

DOI: 10.1021/acs.jmedchem.6b01837 J. Med. Chem. 2017, 60, 5955−5968

Journal of Medicinal Chemistry

Perspective

MPO, multiparameter optimization; pIC50, −log10(IC50); PMI, process mass intensity



(20) Baell, J. B. Broad coverage of commercially available lead-like screening space with fewer than 350,000 compounds. J. Chem. Inf. Model. 2013, 53, 39−55. (21) Kellenberger, E.; Springael, J.-Y.; Parmentier, M.; Hachet-Haas, M.; Galzi, J.-L.; Rognan, D. Identification of nonpeptide CCR5 receptor agonists by structure-based virtual screening. J. Med. Chem. 2007, 50, 1294−1303. (22) Ryckmans, T.; Edwards, M. P.; Horne, V. A.; Correia, A. M.; Owen, D. R.; Thompson, L. R.; Tran, I.; Tutt, M. F.; Young, T. Rapid assessment of a novel series of selective CB2 agonists using parallel synthesis protocols: a lipophilic efficiency (LipE) analysis. Bioorg. Med. Chem. Lett. 2009, 19, 4406−4409. (23) Stumpfe, D.; Bajorath, J. Exploring activity cliffs in medicinal chemistry. J. Med. Chem. 2012, 55, 2932−2942. (24) Cherkasov, A.; Muratov, E. N.; Fourches, D.; Varnek, A.; Baskin, I. I.; Cronin, M.; Dearden, J.; Gramatica, P.; Martin, Y. C.; Todeschini, R.; Consonni, V.; Kuz’min, V. E.; Cramer, R.; Benigni, R.; Yang, C.; Rathman, J.; Terfloth, L.; Gasteiger, J.; Richard, A.; Tropsha, A. QSAR modeling: where have you been? Where are you going to? J. Med. Chem. 2014, 57, 4977−5010. (25) Gunaydin, H. Probabilistic approach to generating MPOs and its application as a scoring function for CNS drugs. ACS Med. Chem. Lett. 2016, 7, 89−93. (26) Segall, M.; Champness, E.; Leeding, C.; Lilien, R.; Mettu, R.; Stevens, B. Applying medicinal chemistry transformations and multiparameter optimization to guide the search for high-quality leads and candidates. J. Chem. Inf. Model. 2011, 51, 2967−2976. (27) Wager, T. T.; Hou, X.; Verhoest, P. R.; Villalobos, A. Moving beyond rules: the development of a central nervous system multiparameter optimization (CNS MPO) approach to enable alignment of druglike properties. ACS Chem. Neurosci. 2010, 1, 435−449. (28) Burke, M. D.; Schreiber, S. L. A planning strategy for diversityoriented synthesis. Angew. Chem., Int. Ed. 2004, 43, 46−58. (29) Ertl, P.; Schuffenhauer, A. Estimation of synthetic accessibility score of drug-like molecules based on molecular complexity and fragment contributions. J. Cheminf. 2009, 1, 8. (30) Brown, D. G.; Boström, J. Analysis of past and present synthetic methodologies on medicinal chemistry: Where have all the new reactions gone? J. Med. Chem. 2016, 59, 4443−4458. (31) Law, J.; Zsoldos, Z.; Simon, A.; Reid, D.; Liu, Y.; Khew, S. Y.; Johnson, A. P.; Major, S.; Wade, R. A.; Ando, H. Y. Route designer: a retrosynthetic analysis tool utilizing automated retrosynthetic rule generation. J. Chem. Inf. Model. 2009, 49, 593−602. (32) Gasteiger, J. Chemoinformatics: achievements and challenges, a personal view. Molecules 2016, 21, 151. (33) Varnek, A.; Baskin, I. I. Chemoinformatics as a theoretical chemistry discipline. Mol. Inf. 2011, 30, 20−32. (34) Kenny, P. W.; Leitão, A.; Montanari, C. A. Ligand efficiency metrics considered harmful. J. Comput.-Aided Mol. Des. 2014, 28, 699− 710. (35) Waring, M. J.; Arrowsmith, J.; Leach, A. R.; Leeson, P. D.; Mandrell, S.; Owen, R. M.; Pairaudeau, G.; Pennie, W. D.; Pickett, S. D.; Wang, J.; Wallace, O.; Weir, A. An analysis of the attrition of drug candidates from four major pharmaceutical companies. Nat. Rev. Drug Discovery 2015, 14, 475−486. (36) Hopkins, A. L.; Keseru, G. M.; Leeson, P. D.; Rees, D. C.; Reynolds, C. H. The role of ligand efficiency metrics in drug discovery. Nat. Rev. Drug Discovery 2014, 13, 105−121. (37) Shultz, M. D. Setting expectations in molecular optimizations: Strengths and limitations of commonly used composite parameters. Bioorg. Med. Chem. Lett. 2013, 23, 5980−5991. (38) Leeson, P. D.; Springthorpe, B. The influence of drug-like concepts on decision-making in medicinal chemistry. Nat. Rev. Drug Discovery 2007, 6, 881−890. (39) Colmenarejo, G. Compound prioritization in single-concentration screening data using ligand efficiency indexes. J. Chem. Inf. Model. 2016, 56, 1705−1713.

REFERENCES

(1) Constable, D. J. C.; Jimenez-Gonzalez, C.; Henderson, R. K. Perspective on solvent use in the pharmaceutical industry. Org. Process Res. Dev. 2007, 11, 133−137. (2) Summerton, L.; Sneddon, H. F.; Jones, L. C.; Clark, J. H. Green and Sustainable Medicinal Chemistry: Methods, Tools and Strategies for the 21st Century Pharmaceutical Industry; The Royal Society of Chemistry: Cambridge, U.K., 2016; pp 1−221. (3) Henderson, R. K.; Jimenez-Gonzalez, C.; Constable, D. J. C.; Alston, S. R.; Inglis, G. G. A.; Fisher, G.; Sherwood, J.; Binks, S. P.; Curzons, A. D. Expanding GSK’s solvent selection guide - embedding sustainability into solvent selection starting at medicinal chemistry. Green Chem. 2011, 13, 854−862. (4) Jimenez-Gonzalez, C.; Ponder, C. S.; Broxterman, Q. B.; Manley, J. B. Using the right green yardstick: why process mass intensity is used in the pharmaceutical industry to drive more sustainable processes. Org. Process Res. Dev. 2011, 15, 912−917. (5) Kjell, D. P.; Watson, I. A.; Wolfe, C. N.; Spitler, J. T. Complexitybased metric for process mass intensity in the pharmaceutical industry. Org. Process Res. Dev. 2013, 17, 169−174. (6) Tucker, J. L. Green chemistry, a pharmaceutical perspective. Org. Process Res. Dev. 2006, 10, 315−319. (7) Anastas, P. T.; Warner, J. C. Green Chemistry: Theory and Practice; Oxford University Press: Oxford, U.K., 1998; p 30. (8) Dunn, P. J.; Wells, A.; Williams, M. T. Green Chemistry in the Pharmaceutical Industry; Wiley-VCH: Weinheim, Germany, 2010; pp 10−352. (9) Sneddon, H. F. Green Chemistry Strategies for Drug Discovery; The Royal Society of Chemistry: Cambridge, U.K., 2015; Vol. 46, p 14. (10) Abou-Gharbia, M.; Childers, W. E. Discovery of innovative therapeutics: today’s realities and tomorrow’s vision. 2. Pharma’s challenges and their commitment to innovation. J. Med. Chem. 2014, 57, 5525−5553. (11) Bryan, M. C.; Dillon, B.; Hamann, L. G.; Hughes, G. J.; Kopach, M. E.; Peterson, E. A.; Pourashraf, M.; Raheem, I.; Richardson, P.; Richter, D.; Sneddon, H. F. Sustainable practices in medicinal chemistry: current state and future directions. J. Med. Chem. 2013, 56, 6007−6021. (12) Tropsha, A.; Bajorath, J. Computational methods for drug discovery and design. J. Med. Chem. 2016, 59, 1−1. (13) Sheridan, R. P.; Zorn, N.; Sherer, E. C.; Campeau, L.-C.; Chang, C.; Cumming, J.; Maddess, M. L.; Nantermet, P. G.; Sinz, C. J.; O’Shea, P. D. Modeling a crowdsourced definition of molecular complexity. J. Chem. Inf. Model. 2014, 54, 1604−1616. (14) Zhou, J. Z. Structure-directed combinatorial library design. Curr. Opin. Chem. Biol. 2008, 12, 379−385. (15) Meanwell, N. A. Improving drug candidates by design: a focus on physicochemical properties as a means of improving compound disposition and safety. Chem. Res. Toxicol. 2011, 24, 1420−1456. (16) Ortwine, D. F.; Aliagas, I. Physicochemical and DMPK in silico models: facilitating their use by medicinal chemists. Mol. Pharmaceutics 2013, 10, 1153−1161. (17) Feng, J. A.; Aliagas, I.; Bergeron, P.; Blaney, J. M.; Bradley, E. K.; Koehler, M. F. T.; Lee, M.-L.; Ortwine, D. F.; Tsui, V.; Wu, J.; Gobbi, A. An integrated suite of modeling tools that empower scientists in structure- and property-based drug design. J. Comput.-Aided Mol. Des. 2015, 29, 511−523. (18) Rankovic, Z. CNS drug design: balancing physicochemical properties for optimal brain exposure. J. Med. Chem. 2015, 58, 2584− 2608. (19) Sheridan, R. P.; McMasters, D. R.; Voigt, J. H.; Wildey, M. J. eCounterscreening: using QSAR predictions to prioritize testing for off-target activities and setting the balance between benefit and risk. J. Chem. Inf. Model. 2015, 55, 231−238. 5966

DOI: 10.1021/acs.jmedchem.6b01837 J. Med. Chem. 2017, 60, 5955−5968

Journal of Medicinal Chemistry

Perspective

(40) Holenz, J. Lead Generation: Methods and Strategies; Wiley-VCH: Weinheim, Germany, 2016; Vol. 68, pp 451−462. (41) Tanaka, D.; Tsuda, Y.; Shiyama, T.; Nishimura, T.; Chiyo, N.; Tominaga, Y.; Sawada, N.; Mimoto, T.; Kusunose, N. A practical use of ligand efficiency indices out of the fragment-based approach: ligand efficiency-guided lead identification of soluble epoxide hydrolase inhibitors. J. Med. Chem. 2011, 54, 851−857. (42) Tarcsay, Á .; Nyíri, K.; Keserű , G. M. Impact of lipophilic efficiency on compound quality. J. Med. Chem. 2012, 55, 1252−1260. (43) Segall, M. D. Multi-parameter optimization: identifying high quality compounds with a balance of properties. Curr. Pharm. Des. 2012, 18, 1292−1310. (44) de la Vega de Leon, A.; Bajorath, J. Matched molecular pairs derived by retrosynthetic fragmentation. MedChemComm 2014, 5, 64− 67. (45) Griffen, E.; Leach, A. G.; Robb, G. R.; Warner, D. J. Matched molecular pairs as a medicinal chemistry tool. J. Med. Chem. 2011, 54, 7739−7750. (46) Hu, X.; Hu, Y.; Vogt, M.; Stumpfe, D.; Bajorath, J. MMP-Cliffs: systematic identification of activity cliffs on the basis of matched molecular pairs. J. Chem. Inf. Model. 2012, 52, 1138−1145. (47) Sushko, Y.; Novotarskyi, S.; Körner, R.; Vogt, J.; Abdelaziz, A.; Tetko, I. V. Prediction-driven matched molecular pairs to interpret QSARs and aid the molecular optimization process. J. Cheminf. 2014, 6, 48. (48) Fernandez, D.; Ghanta, A.; Kauffman, G. W.; Sanguinetti, M. C. Physicochemical features of the hERG channel drug binding site. J. Biol. Chem. 2004, 279, 10120−10127. (49) Cummins, D. J.; Bell, M. A. Integrating everything: the molecule selection toolkit, a system for compound prioritization in drug discovery. J. Med. Chem. 2016, 59, 6999−7010. (50) Zhou, J. Z. Chemical Library Design; Humana Press: New York, 2011; pp 3−154. (51) Xing, L.; McDonald, J. J.; Kolodziej, S. A.; Kurumbail, R. G.; Williams, J. M.; Warren, C. J.; O’Neal, J. M.; Skepner, J. E.; Roberds, S. L. Discovery of potent inhibitors of soluble epoxide hydrolase by combinatorial library design and structure-based virtual screening. J. Med. Chem. 2011, 54, 1211−1222. (52) Liao, C.; Sitzmann, M.; Pugliese, A.; Nicklaus, M. C. Software and resources for computational medicinal chemistry. Future Med. Chem. 2011, 3, 1057−1085. (53) Hu, Q.; Peng, Z.; Sutton, S. C.; Na, J.; Kostrowicki, J.; Yang, B.; Thacher, T.; Kong, X.; Mattaparti, S.; Zhou, J. Z.; Gonzalez, J.; Ramirez-Weinhouse, M.; Kuki, A. Pfizer global virtual library (PGVL): A chemistry design tool powered by experimentally validated parallel synthesis information. ACS Comb. Sci. 2012, 14, 579−589. (54) Aliagas, I.; Gobbi, A.; Heffron, T.; Lee, M.-L.; Ortwine, D. F.; Zak, M.; Khojasteh, S. C. A probabilistic method to report predictions from a human liver microsomes stability QSAR model: a practical tool for drug discovery. J. Comput.-Aided Mol. Des. 2015, 29, 327−338. (55) Lee, M.-L.; Aliagas, I.; Dotson, J.; Feng, J. A.; Gobbi, A.; Heffron, T. DEGAS: Sharing and tracking target compound ideas with external collaborators. J. Chem. Inf. Model. 2012, 52, 278−284. (56) Ritchie, T. J.; Macdonald, S. J. F. The impact of aromatic ring count on compound developability − are too many aromatic rings a liability in drug design? Drug Discovery Today 2009, 14, 1011−1020. (57) Skerratt, S. E.; Andrews, M.; Bagal, S. K.; Bilsland, J.; Brown, D.; Bungay, P. J.; Cole, S.; Gibson, K. R.; Jones, R.; Morao, I.; Nedderman, A.; Omoto, K.; Robinson, C.; Ryckmans, T.; Skinner, K.; Stupple, P.; Waldron, G. The discovery of a potent, selective, and peripherally restricted pan-Trk inhibitor (PF-06273340) for the treatment of pain. J. Med. Chem. 2016, 59, 10084−10099. (58) Kulagowski, J. J.; Blair, W.; Bull, R. J.; Chang, C.; Deshmukh, G.; Dyke, H. J.; Eigenbrot, C.; Ghilardi, N.; Gibbons, P.; Harrison, T. K.; Hewitt, P. R.; Liimatta, M.; Hurley, C. A.; Johnson, A.; Johnson, T.; Kenny, J. R.; Bir Kohli, P.; Maxey, R. J.; Mendonca, R.; Mortara, K.; Murray, J.; Narukulla, R.; Shia, S.; Steffek, M.; Ubhayakar, S.; Ultsch, M.; van Abbema, A.; Ward, S. I.; Waszkowycz, B.; Zak, M.

Identification of imidazo-pyrrolopyridines as novel and potent JAK1 inhibitors. J. Med. Chem. 2012, 55, 5901−5921. (59) Zak, M.; Mendonca, R.; Balazs, M.; Barrett, K.; Bergeron, P.; Blair, W. S.; Chang, C.; Deshmukh, G.; DeVoss, J.; Dragovich, P. S.; Eigenbrot, C.; Ghilardi, N.; Gibbons, P.; Gradl, S.; Hamman, C.; Hanan, E. J.; Harstad, E.; Hewitt, P. R.; Hurley, C. A.; Jin, T.; Johnson, A.; Johnson, T.; Kenny, J. R.; Koehler, M. F. T.; Bir Kohli, P.; Kulagowski, J. J.; Labadie, S.; Liao, J.; Liimatta, M.; Lin, Z.; Lupardus, P. J.; Maxey, R. J.; Murray, J. M.; Pulk, R.; Rodriguez, M.; Savage, S.; Shia, S.; Steffek, M.; Ubhayakar, S.; Ultsch, M.; van Abbema, A.; Ward, S. I.; Xiao, L.; Xiao, Y. Discovery and optimization of C-2 methyl imidazopyrrolopyridines as potent and orally bioavailable JAK1 inhibitors with selectivity over JAK2. J. Med. Chem. 2012, 55, 6176− 6193. (60) Gao, H.; Shanmugasundaram, V.; Lee, P. Estimation of aqueous solubility of organic compounds with QSPR approach. Pharm. Res. 2002, 19, 497−503. (61) Lee, P. H.; Cucurull-Sanchez, L.; Lu, J.; Du, Y. J. Development of in silico models for human liver microsomal stability. J. Comput.Aided Mol. Des. 2007, 21, 665−673. (62) Birch, A. M.; Birtles, S.; Buckett, L. K.; Kemmitt, P. D.; Smith, G. J.; Smith, T. J. D.; Turnbull, A. V.; Wang, S. J. Y. Discovery of a potent, selective, and orally efficacious pyrimidinooxazinyl bicyclooctaneacetic acid diacylglycerol acyltransferase-1 inhibitor. J. Med. Chem. 2009, 52, 1558−1568. (63) Galloway, W. R. J. D.; Spring, D. R. Towards drugging the ‘undruggable’: enhancing the scaffold diversity of synthetic small molecule screening collections using diversity-oriented synthesis. Diversity-Oriented Synth. 2013, 1, 21−28. (64) Ibbeson, B. M.; Laraia, L.; Alza, E.; O’ Connor, C. J.; Tan, Y. S.; Davies, H. M. L.; McKenzie, G.; Venkitaraman, A. R.; Spring, D. R. Diversity-oriented synthesis as a tool for identifying new modulators of mitosis. Nat. Commun. 2014, 5, 3155. (65) O’ Connor, C. J.; Beckmann, H. S. G.; Spring, D. R. Diversityoriented synthesis: producing chemical tools for dissecting biology. Chem. Soc. Rev. 2012, 41, 4444−4456. (66) Rajarathinam, B.; Kumaravel, K.; Vasuki, G. Green chemistry oriented multi-component strategy to hybrid heterocycles. RSC Adv. 2016, 6, 73848−73852. (67) Tan, D. S. Diversity-oriented synthesis: exploring the intersections between chemistry and biology. Nat. Chem. Biol. 2005, 1, 74−84. (68) Marcaurelle, L. A.; Comer, E.; Dandapani, S.; Duvall, J. R.; Gerard, B.; Kesavan, S.; Lee, M. D.; Liu, H.; Lowe, J. T.; Marie, J.-C.; Mulrooney, C. A.; Pandya, B. A.; Rowley, A.; Ryba, T. D.; Suh, B.-C.; Wei, J.; Young, D. W.; Akella, L. B.; Ross, N. T.; Zhang, Y.-L.; Fass, D. M.; Reis, S. A.; Zhao, W.-N.; Haggarty, S. J.; Palmer, M.; Foley, M. A. An aldol-based build/couple/pair strategy for the synthesis of medium- and large-sized rings: discovery of macrocyclic histone deacetylase inhibitors. J. Am. Chem. Soc. 2010, 132, 16962−16976. (69) Comer, E.; Beaudoin, J. A.; Kato, N.; Fitzgerald, M. E.; Heidebrecht, R. W.; Lee, M. d.; Masi, D.; Mercier, M.; Mulrooney, C.; Muncipinto, G.; Rowley, A.; Crespo-Llado, K.; Serrano, A. E.; Lukens, A. K.; Wiegand, R. C.; Wirth, D. F.; Palmer, M. A.; Foley, M. A.; Munoz, B.; Scherer, C. A.; Duvall, J. R.; Schreiber, S. L. Diversityoriented synthesis-facilitated medicinal chemistry: toward the development of novel antimalarial agents. J. Med. Chem. 2014, 57, 8496−8502. (70) Kato, N.; Comer, E.; Sakata-Kato, T.; Sharma, A.; Sharma, M.; Maetani, M.; Bastien, J.; Brancucci, N. M.; Bittker, J. A.; Corey, V.; Clarke, D.; Derbyshire, E. R.; Dornan, G. L.; Duffy, S.; Eckley, S.; Itoe, M. A.; Koolen, K. M. J.; Lewis, T. A.; Lui, P. S.; Lukens, A. K.; Lund, E.; March, S.; Meibalan, E.; Meier, B. C.; McPhail, J. A.; Mitasev, B.; Moss, E. L.; Sayes, M.; Van Gessel, Y.; Wawer, M. J.; Yoshinaga, T.; Zeeman, A.-M.; Avery, V. M.; Bhatia, S. N.; Burke, J. E.; Catteruccia, F.; Clardy, J. C.; Clemons, P. A.; Dechering, K. J.; Duvall, J. R.; Foley, M. A.; Gusovsky, F.; Kocken, C. H. M.; Marti, M.; Morningstar, M. L.; Munoz, B.; Neafsey, D. E.; Sharma, A.; Winzeler, E. A.; Wirth, D. F.; Scherer, C. A.; Schreiber, S. L. Diversity-oriented synthesis yields novel multistage antimalarial inhibitors. Nature 2016, 538, 344−349. 5967

DOI: 10.1021/acs.jmedchem.6b01837 J. Med. Chem. 2017, 60, 5955−5968

Journal of Medicinal Chemistry

Perspective

Stavrinides, G. S., Banerjee, S., Caglar, H. S., Ozer, M., Eds.; Springer: Berlin, 2013; pp 301−306. (90) Whitlock, H. W. On the structure of total synthesis of complex natural products. J. Org. Chem. 1998, 63, 7982−7989. (91) Hack, M. D.; Rassokhin, D. N.; Buyck, C.; Seierstad, M.; Skalkin, A.; ten Holte, P.; Jones, T. K.; Mirzadegan, T.; Agrafiotis, D. K. Library enhancement through the wisdom of crowds. J. Chem. Inf. Model. 2011, 51, 3275−3286. (92) Rastogi, T.; Leder, C.; Kümmerer, K. Re-designing of existing pharmaceuticals for environmental biodegradability: a tiered approach with β-blocker propranolol as an example. Environ. Sci. Technol. 2015, 49, 11756−11763.

(71) Dandapani, S.; Germain, A. R.; Jewett, I.; le Quement, S.; Marie, J.-C.; Muncipinto, G.; Duvall, J. R.; Carmody, L. C.; Perez, J. R.; Engel, J. C.; Gut, J.; Kellar, D.; Siqueira-Neto, J. L.; McKerrow, J. H.; Kaiser, M.; Rodriguez, A.; Palmer, M. A.; Foley, M.; Schreiber, S. L.; Munoz, B. Diversity-oriented synthesis yields a new drug lead for treatment of chagas disease. ACS Med. Chem. Lett. 2014, 5, 149−153. (72) Ganesan, A.; Coote, M. L.; Barakat, K. Molecular dynamicsdriven drug discovery: leaping forward with confidence. Drug Discovery Today 2017, 22, 249−269. (73) Ooms, F. Molecular modeling and computer aided drug design. examples of their applications in medicinal chemistry. Curr. Med. Chem. 2000, 7, 141−158. (74) Evenäs, J.; Edfeldt, F.; Lepistö, M.; Svitacheva, N.; Synnergren, A.; Lundquist, B.; Gränse, M.; Rönnholm, A.; Varga, M.; Wright, J.; Wei, M.; Yue, S.; Wang, J.; Li, C.; Li, X.; Chen, G.; Liao, Y.; Lv, G.; Tjörnebo, A.; Narjes, F. HTS followed by NMR based counterscreening. Discovery and optimization of pyrimidones as reversible and competitive inhibitors of xanthine oxidase. Bioorg. Med. Chem. Lett. 2014, 24, 1315−1321. (75) LaPlante, S. R.; Forgione, P.; Boucher, C.; Coulombe, R.; Gillard, J.; Hucke, O.; Jakalian, A.; Joly, M.-A.; Kukolj, G.; Lemke, C.; McCollum, R.; Titolo, S.; Beaulieu, P. L.; Stammers, T. Enantiomeric atropisomers inhibit HCV polymerase and/or HIV matrix: characterizing hindered bond rotations and target selectivity. J. Med. Chem. 2014, 57, 1944−1951. (76) Kettle, J. G.; Alwan, H.; Bista, M.; Breed, J.; Davies, N. L.; Eckersley, K.; Fillery, S.; Foote, K. M.; Goodwin, L.; Jones, D. R.; Käck, H.; Lau, A.; Nissink, J. W. M.; Read, J.; Scott, J. S.; Taylor, B.; Walker, G.; Wissler, L.; Wylot, M. Potent and selective inhibitors of MTH1 probe its role in cancer cell survival. J. Med. Chem. 2016, 59, 2346−2361. (77) Cernak, T.; Dykstra, K. D.; Tyagarajan, S.; Vachal, P.; Krska, S. W. The medicinal chemist’s toolbox for late stage functionalization of drug-like molecules. Chem. Soc. Rev. 2016, 45, 546−576. (78) Sherer, E. C.; Lee, C. H.; Shpungin, J.; Cuff, J. F.; Da, C.; Ball, R.; Bach, R.; Crespo, A.; Gong, X.; Welch, C. J. Systematic approach to conformational sampling for assigning absolute configuration using vibrational circular dichroism. J. Med. Chem. 2014, 57, 477−494. (79) Rücker, C.; Rücker, G.; Bertz, S. H. Organic synthesis − art or science? J. Chem. Inf. Comput. Sci. 2004, 44, 378−386. (80) Li, J.; Eastgate, M. D. Current complexity: a tool for assessing the complexity of organic molecules. Org. Biomol. Chem. 2015, 13, 7164−7176. (81) Allu, T. K.; Oprea, T. I. Rapid evaluation of synthetic and molecular complexity for in silico chemistry. J. Chem. Inf. Model. 2005, 45, 1237−1243. (82) Barone, R.; Chanon, M. A new and simple approach to chemical complexity. application to the synthesis of natural products. J. Chem. Inf. Comput. Sci. 2001, 41, 269−272. (83) Bertz, S. H. The first general index of molecular complexity. J. Am. Chem. Soc. 1981, 103, 3599−3601. (84) Bertz, S. H. Complexity of synthetic reactions. The use of complexity indices to evaluate reactions, transforms and disconnections. New J. Chem. 2003, 27, 860−869. (85) Erlanson, D. A.; Fesik, S. W.; Hubbard, R. E.; Jahnke, W.; Jhoti, H. Twenty years on: the impact of fragments on drug discovery. Nat. Rev. Drug Discovery 2016, 15, 605−619. (86) Hendrickson, J. B.; Huang, P.; Toczko, A. G. Molecular complexity: a simplified formula adapted to individual atoms. J. Chem. Inf. Model. 1987, 27, 63−67. (87) Leach, A. R.; Hann, M. M. Molecular complexity and fragmentbased drug discovery: ten years on. Curr. Opin. Chem. Biol. 2011, 15, 489−496. (88) Schuffenhauer, A.; Brown, N.; Selzer, P.; Ertl, P.; Jacoby, E. Relationships between molecular complexity, biological activity, and structural diversity. J. Chem. Inf. Model. 2006, 46, 525−535. (89) von Korff, M.; Sander, T. In Chaos and Complex Systems, Proceedings of the 4th International Interdisciplinary Chaos Symposium; 5968

DOI: 10.1021/acs.jmedchem.6b01837 J. Med. Chem. 2017, 60, 5955−5968