Copyright 2008 by the American Chemical Society Volume 51, Number 13 Recent Developments in Fragment-Based Drug Discovery
Miles Congreve,* Gianni Chessari, Dominic Tisi, and Andrew J. Woodhead Astex Therapeutics Ltd., 436 Cambridge Science Park, Milton Road, Cambridge CB4 0QA, U.K. ReceiVed January 15, 2008 the screening techniques employed in FBDD must be cor- The field of fragment-based drug discovery (FBDDa) has respondingly much more sensitive than a HTS bioassay.
developed significantly over the past 10 years and is now Generally, sensitive biophysical techniques are employed to recognized as a tangible alternative to more traditional methods detect these weak binding events and to characterize the of hit identification, such as high throughput screening (HTS).
fragment interactions with the target active site. Nuclear The number of commercial and academic groups actively magnetic resonance (NMR) and protein X-ray crystallography engaged in fragment-based research has increased, and as a have been used extensively in fragment-based research because consequence, there has been continued development and refine- these techniques are highly sensitive in detecting low affinity ment of techniques and methods. From its inception, the fragment binding and also give information about the fragment- fragment-based approach had two central tenets that were critical protein interactions being formed.
to its success and that have set it apart from HTS and other hit There have been a number of recently published general identification techniques. The first is the concept that chemical review articles that have discussed the various aspects of the space can be more efficiently probed by screening collections FBDD field.3–21 In addition, there are now two books on the of small fragments rather than libraries of larger molecules. The subject.22,23 In this journal in 2004, Erlanson et al. summarized number of potential fragments with up to 12 heavy atoms (not the major developments in FBDD since the original publication including three- and four-membered ring structures) has been by Fesik and co-workers of the "SAR by NMR" approach in estimated at 107,1 whereas the number of potential druglike the late 1990s.3,24 Particular note was given to the biophysical molecules with up to 30 heavy atoms is estimated at more than methods employed to screen for fragment binding and the merits 1060.2 Therefore, a much greater proportion of "fragment-like" and drawbacks of each of these techniques, along with the chemical space can feasibly be screened in FBDD compared to approaches that can be used to optimize fragments into lead "druglike" chemical space covered in a HTS where molecular molecules. Herein, the trends and developments over the past size is much larger. The second idea is that, because by 4 years will be outlined and some selected examples that are definition fragment molecules are small in size (typically less illustrative of the approaches being utilized by those active in than 250 Da), they should typically bind with lower affinity to the field examined. Additionally, this review will look in some their target protein (micromolar to millimolar range) compared detail at representative protein-ligand complexes observed with druglike molecules that can form many more interactions between fragment-sized molecules and their protein targets from (nanomolar to micromolar range) but that the binding efficiency the Protein Data Bank (PDB). Finally, some conclusions will per atom is at least as high as for larger hit molecules. Implicitly, be drawn from these data and the future of FBDD discussed.
* To whom correspondence should be addressed. Phone: +44 (0)1223 226270. Fax +44 (0)1223 226201. E-mail: m.congreve@astex-therapeu- 2. Trends and Developments
a Abbreviations: FBDD, fragment-based drug discovery; HTS, high- In the first part of this review we will examine how things throughput screening; HCS, high-concentration screening; LE, ligand have evolved and developed in the field of FBDD over the past efficiency; HAC, heavy atom count; GE, group efficiency; SBDD, structure-based drug design; MWT, molecular weight; PDB, Protein Data Bank.
10.1021/jm8000373 CCC: $40.75  2008 American Chemical Society Published on Web 05/06/2008 Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 2.1. Fragment Screening. Fragment-based screening has an
the simplest ligand in a drug discovery campaign has influenced intuitive appeal. The success of pharmaceutical companies like those involved in HTS. There have been a number of reports Abbott and biotechnology companies such as Astex Therapeu- of the use of high-concentration screening (HCS) or "reduced tics, SGX Pharmaceuticals, and Plexxikon in developing frag- complexity" screening on compound collections that are a hybrid ments into clinical candidates has influenced the chemistry of a true fragment library and of a typical HTS collection.12,43–45 community, prompting fragment screening efforts in many other In some cases, a somewhat looser set of criteria are applied industrial and academic institutions.16 when constructing fragment libraries than would be used for a In industry over the past 3-4 years, a great deal of effort fragment set designed for biophysical screening (for example, has been given to establishing fragment-based screening, and by allowing the upper molecular weight to approach 350 Da).
it is now generally being implemented as a complementary This has the effect of greatly increasing the available subset of strategy to HTS. This is in some part due to the fact that molecules that can be screened within a corporate collection.
investments made during the 1990s in HTS and combinatorial Screening larger molecules (that are capable of forming more chemistry have not yielded success for more challenging classes interactions with the target protein and hence delivering higher of drug targets. However, despite the obvious efforts to potency) has the additional benefit that they will, if active, be implement fragment screening, there are significant cultural and detectable in a HTS campaign simply by screening at a higher practical issues to overcome within large companies to apply than normal concentration.45 This strategy is essentially identical this new methodology in an effective manner. In particular, after to that suggested in 1999 by Teague et al. in which the authors identification of fragment hits, optimization to a more conven- argue that "leadlike" hits (<350 Da) are advantageous as start tional potency range will often be difficult without structural points for hit-to-lead chemistry compared with the "drug-sized" information. Significant up-front investment in structural biology hits typically delivered from HTS, even if of lower potency.46,47 is required both to establish the binding modes of fragments Having a looser set of criteria for fragment libraries does, within the active site of target proteins and to eliminate any however, raise a number of potential issues, and these are false positives. This commitment to timely structural biology discussed later.
may be difficult to achieve in practice in large organizations, An additional trend, often when there is an absence of particularly when only a proportion of targets are readily structural information or when there is limited chemistry amenable to 3D-structure determination. Another issue is that resource to follow up "primary" fragment hits, is to carry out a fragment hits with low or undetectable potency in a biological secondary screen of leadlike analogues in a conventional assay may initially appear less attractive to medicinal chemists bioassay. This iteration of SAR generation can lead to identi- when compared with conventional HTS hits with higher potency.
fication of analogues with potencies similar to those of typical In contrast, in an academic setting assembling a small library HTS hits, ensuring that the fragment derived hits are competitive of fragments and screening using a biophysical technique such with those emerging from a standard screen and giving as surface plasmon resonance (SPR), protein-ligand NMR, or confidence in the results from the fragment screening. For even X-ray crystallography is much more achievable compared example, at Vertex the NMR SHAPES method has been used with assembling and screening a large library in a bioassay. In successfully for 10 years as a complementary approach to HTS fact, some of the pioneering work using X-ray crystallography and typically 500 follow-up compounds are screened around a for fragment screening was done at the University of Groningen promising fragment hit, often leading to the discovery of hits in the early 1990s.25 State of the art high field NMR instruments with 5-10 µM potency (70-80% success rate).48 and expertise in other biophysical techniques will often be 2.2. Ligand Efficiency. A valuable concept now widely used
readily available within world class academic groups, and there for comparing hits across different series and the effectiveness have been a number of recent publications from academia in of compound optimization is ligand efficiency (LE). The term which FBDD methods have been applied to exploratory LE is defined as the free energy of binding of a ligand for a specific protein averaged for each "heavy atom" (or non-hydrogen atom).49–51 The number of heavy atoms is termed the Since FBDD was last reviewed in this journal in 2004, there heavy atom count (HAC).
has been continued development of fragment-based screeningtechnologies, bringing higher throughput, increased cost ef- LE ) -∆G ⁄ HAC ≈ -RT ln(IC ) ⁄ HAC fectiveness, and reductions in protein requirements as well as potentially broadening the application to a wider range of Throughout this review the units of LE are (kcal/mol)/heavy therapeutic targets. In this review we will not cover these atom. If we then consider that an oral drug candidate should improved methods in any detail but will briefly outline some have a molecular weight of <500 Da (to fulfill Lipinski's rules) of the new technological advances. One important development and typically IC < 10 nM, we can extrapolate that a minimum is that the cryogenic NMR probe has become much more widely LE of 0.3 is required in a hit or lead for it to be useful. This available, giving greatly improved sensitivity and therefore value can then be used as a guideline for the LE required in a improving data quality and throughput of NMR-based screening good fragment, assuming LE cannot be improved during the approaches.34 Another example is the use of NMR to detect the displacement of "spy" molecules that contain 19F from Kuntz et al. in 1999 were the first to calculate normalized protein active sites.35 The introduction of cryogenic probe potencies in a similar way and also to suggest that such values technology with 19F detection capabilities or the use of capillary would be useful in tracking the potencies of molecules.52 From NMR probes36 may in future increase the sensitivity of this the outset, groups carrying out fragment-based discovery were method to levels that rival traditional enzymatic assays. We have generally monitoring potency vs molecular weight during also seen further refinements in protein-ligand X-ray crystal- fragment optimization, but this idea was not more widely lography,37 mass spectrometry,38,39 surface plasmon resonance,40–42 adopted until Hopkins et al. introduced the term LE in 2004.49 and isothermal titration calorimetry.33 Today, not only is it used by practitioners of fragment-based As well as an increase in the number of groups using drug discovery,9,51,53 but it is being adopted more generally by fragment-based screening methods, the concept of starting from the medicinal chemistry community. The use of LE helps to Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 3663
concluded that more lipophilic compounds carry significant riskswith them into development, especially with respect to increasedchances of nonspecific toxicity.55 It was noted that manypharmaceutical companies are making molecules with relativelyhigh cLogP values (based on the patent literature). They havetherefore proposed the use of a new index called "ligand-lipophilicity efficiency" (LLE) as an important guide during leadoptimization and as a flag for preclinical candidate selection.
LLE is defined as LLE ) pIC50 (or pKi) - cLogP (or LogD),and the target LLE for a low nanomolar potency lead was Figure 1. Group efficiencies of a protein kinase B inhibitor and
suggested as ∼5-7 or greater. Fragment hits tend to be very fragment potencies used to derive the GE for the methyl group at the polar and water soluble, and the experience at Astex is that the 5-position of the pyrazole ring.
compounds synthesized in order to optimize fragments tend to reduce the seductive influence of increased potency after each have lower molecular weight and higher polarity when compared iteration of compound design and serves to remind the chemist with those outlined in Leeson et al.'s article. For example, in that consideration of the physicochemical properties of a lead 2006 the median cLogP value for patented molecules from Astex series is equally important. Recently, modified definitions of was 2.4, significantly lower than for the equivalent values LE have been proposed,50,54 and there is also an increased presented in the paper from some large pharmaceutical com- emphasis on tracking increasing lipophilicity with potency.55 panies (AstraZeneca, 3.7; GlaxoSmithKline, 4.2; Merck, 4.0; Overall, monitoring of LE provides a conceptual "roadmap" Pfizer, 3.5). This may then be an additional advantage of using for the fragment optimization process.51 the fragment-based approach.
2.3. Group Efficiency. A recent evolution of LE is to look
2.5. Fragment Libraries. There have been a number of
at group efficiency (GE). This allows for the estimation of an discussions in the recent literature of what constitutes a good individual group's contribution toward the overall free energy fragment library. The main considerations are (1) the range of of binding (∆G), giving a quick and simple insight into how physicochemical properties of fragments to be included, (2) aqueous efficient one modification is over another. This analysis requires solubility and quality control, (3) assessment of molecular diversity, the comparison of matched pairs of compounds using a (4) chemical tractability of the fragments for follow-up, (5) Free-Wilson analysis. The ∆G values can be converted into which chemical functionalities are disallowed, (6) druglikeness GE in an analogous manner to LE, GE ) -∆G/HAC, where of the fragments with respect to precedence of the templates in HAC is the number of non-hydrogen atoms in a particular group.
oral drugs and natural products, and (7) sampling of privileged GE g 0.3 indicates that the group is making an acceptable medicinal chemistry scaffolds.4,12,19,20,37,45,58–60 Although there contribution to the compound's potency overall, ensuring are clearly a number of "flavors" of fragment libraries emerging maintenance of good druglike properties.28,56 Figure 1 shows and slightly different definitions of what constitutes a good the group efficiencies for various parts of a potent protein kinase fragment, these reports are broadly in agreement. In simple B (PKB/Akt) inhibitor and effectively highlights key hot spots terms, a fragment should have a maximum molecular weight for binding with this type of molecule. Matched compounds of 300 Da (as proposed originally by Astex Therapeutics in its were used to determine the GE for various parts of the molecule; "rule of three"), some complexity filters should be applied (either for example, fragments A and B were used to determine the based on 2D descriptors such as H-bond donors, H-bond GE for the methyl group at the 5-position of the pyrazole acceptors, or rotatable bonds or by using a complexity fingerprint For this approach to be valid, it is important to be sure that approach), and high aqueous solubility is essential for practical the binding modes for the compounds that are being compared reasons during screening.61 are similar; otherwise, the data could become very misleading.
Despite this general level of agreement of what a useful This information can be used to quickly build up which parts fragment library looks like, we saw in the previous section how of the active site are "hot spots" for gaining affinity and which HCS is being used as one approach to screen fragment libraries groups on the inhibitor are the most effective. GE can also be using molecules that tend to be at the upper limits of what would used to elegantly illustrate how the addition of relatively large constitute a fragment. Indeed, it is our belief that HCS will groups with moderate gains in potency is not necessarily helpful become very popular in large pharmaceutical organisations during optimization (for example, the addition of a phenyl group because it requires very little change to current best practice should increase potency by at least 50-fold when GE > 0.3).
and might be expected to deliver high quality hits for more However, this utility comes with the caveat that the underlying tractable targets. However, the two basic principles of FBDD assumption of the group-based additivity of free energies of (the use of very small compounds and very high screening binding is an approximation and will not always be true.57 sensitivity) need to be carefully reconsidered when using HCS 2.4. Ligand-Lipophilicity Efficiency. Another way of as-
for fragment libraries. There are two main issues to be addressed.
sessing "druggability" is to take into account lipophilicity. This First, as the size of a fragment increases, the number of sensible can be done by flagging highly lipophilic compounds that may molecules will grow dramatically, making it much harder to be deriving a significant proportion of their binding affinity from design a diverse library.1 Additionally, it is now generally desolvation. A significant part of protein-ligand binding accepted that as molecular size and complexity grows, the involves desolvation of the ligand, so the more lipophilic the chances of an H-bonding mismatch or steric clash between ligand, the more favorable this process will tend to be. This the fragment and the target protein rapidly increase, reducing implies that lipophilic molecules should have an increased the chance of finding hits.62 Both of these factors suggest that chance of binding to any given drug-sized pocket and that a much larger library of "leadlike" compounds will be required lipophilic compounds should be more promiscuous than polar to achieve a hit rate comparable to that generally observed for ones as a result. Leeson and Springthorpe have carefully small fragments screened using very sensitive biophysical analyzed the effects of lipophilicity on drug failures and techniques at a higher concentration. Second, at lower concen- Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 Figure 2. Comparison of the distributions of commercially available
compounds that are candidates for different types of screening
collection. Twenty-two preferred suppliers were selected. Rule of three
(Ro3): MWT e 300, ClogP e 3, Donors e 3, Acceptors e 3. Leadlike:
Figure 3. Plots of the number of heavy atoms of a given fragment vs
MWT e 350, ClogP e 3 or 5, Donors e 5, Acceptors e 10. Rule of the ligand efficiency that would be required for that compound to be Five (Ro5): MWT e 500, cLogP e 5, Donors e 5, Acceptors e 10.
detected as a hit when screened at 2 mM (blue), 200 µM (purple), and All compounds must be listed as "available" and must contain at least 20 µM (orange).
one carbon atom in a ring. The allowed elements are 1H, C, N, O, S,F, Cl, Br, I.
filters of MWT e 350 and either cLogP e 3 (as originallydescribed by Teague et al.) or cLogP e 5 (according to trations the smallest fragments will only be detectable if they Lipinski's rules) and also requiring compounds to obey the other have potency similar to that of the larger, more complex rules suggested by Lipinski produce libraries of approximately compounds being screened. This means that the individual 400 000 and 700 000 compounds, respectively (lead-like library interactions must be very efficient in the smallest fragments, or cLogP e 3 and leadlike library cLogP e 5), with a right-shifted the binding information will be lost in the noise of the assay.
distribution toward members with higher molecular weight.46 This last factor is captured by the concept of ligand efficiency, Figure 2 illustrates three main points. First, a molecular weight which was discussed earlier.49 cutoff alone is a poor start point for defining a library of In considering the first of these two issues (the idea that molecules because without cLogP and complexity filters the "chemical space" explodes dramatically as molecular size distribution of available molecules will be skewed toward the increases), an assessment of the relative sizes of different upper mass limit. Second, the available Ro3 library is signifi- libraries that might currently be assembled from commercial cantly smaller than the Ro5 and leadlike libraries, suggesting it sources can be made. Although this analysis does not measure should be practically more straightforward to create a library how actual chemical diversity changes with molecular size, it of compounds that adequately represents "commercially avail- should help assess how thoroughly one can sample chemical able diversity" within this area of chemical space than for the space in practice. Figure 2 plots the distribution of four libraries other two types of library. Lastly, even a small increase in the of available molecules that can be purchased from commercial upper molecular weight limit (perhaps 50 Da) in a fragment databases. In each case the compounds considered for possible library will dramatically increase the number of available inclusion contained carbon and at least one ring and did not molecules that can be considered. True chemical diversity will contain inorganic elements or deuterium isotopes. There are actually have massively increased, meaning that the largest many more druglike filters that could additionally be applied, fragments within the library will inevitably represent their region but for the purposes of the comparisons on the chart no further of chemical space much less well than the smallest com- filtering has been used. By use of these simple constraints on a subset of our in-house compound database of commercial To examine the issue of hit detection outlined above, it is samples, 2 255 680 compounds were found to be potentially useful to again consider LE. Figure 3 plots the HAC of a given available. By use of the "rule of three" filters (Ro3 library; MWT fragment vs the LE that would be required for that compound e 300, cLogP e 3, number of H-bond donors e 3, number of to be detected as a hit when screened at 2 mM (dark blue), 200 H-bond acceptors e 3), there are approximately 137 000 µM (purple), and 20 µM (orange). A fragment with a HAC of possible compounds available in a fairly even distribution across 12 (MWT typically 160-170 Da) will require a LE of 0.3 or the 7-23 heavy atom range. This distribution is tightly greater to be detected when screened at 2 mM (a concentration controlled by the restrictions on numbers of donors and that can only be reliably screened using a biophysical method, acceptors, which serves to limit molecular complexity. Similarly, such as protein-ligand NMR, SPR, or protein-ligand crystal- Siegal et al. have suggested that, after applying more rigorous lography). As discussed earlier, this is the minimum useful LE druglike selection filters than used here, 70 000 "Ro3 compliant" required to develop a fragment into a potent molecule that obeys fragments are commercially available from preferred suppliers.63 Lipinski's rules. For the same fragment to be detectable at a By comparision of this with compounds selected using Lipin- concentration of 200 µM (for example, in HCS), it would require ski's "rule of five" guidelines for oral drugs (Ro5 library; MWT a LE of 0.42, and any fragments with lower LE would not be e 500, cLogP e 5, number of H-bond donors e 5, number of detectable. A LE of 0.42 or greater is much less common than H-bond acceptors e 10), there is an even overall distribution a LE of around 0.3 because it generally requires multiple high either side of 26 heavy atoms and a total of approximately quality interactions to be present between the ligand and the 1 700 000 compounds to choose from. Finally, applying leadlike protein and (depending on the difficulty of the target) may not Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 3665
Scheme 1. Fragment Evolution of 6-Propyl Isocytosine Leading to a Potent Nonpeptidic BACE-1 Inhibitor
always be achievable. If the same molecule was screened at 20 Scheme 2. Fragment Evolution of the (R)-Enantiomer of the
Oral Drug Mexiletine to a Potent uPA Inhibitor
µM (in a typical HTS campaign), it would require a LE of 0.53.
Such high LE values are rarely observed and may require theprotein target to be highly tractable (for example, a kinaseenzyme target would be suitable).
The above analysis indicates that fragments with HAC of around 12 or less are unlikely to be detectable in a HTS for allbut the most tractable of drug targets, and even in a HCS it isprobable that many fragments that are binding to the target willnot be doing so efficiently enough to be detected. Furthermore,if we examine the curves, we can see that for very smallfragments (HAC ) 8 or less) in most cases X-ray crystal- found in the literature, and it is not the intention to list these lography will be required to detect their binding, as only this here but rather to select one or two representative and influential approach is likely to be reliable in the 2-10 mM binding range.
case histories that illustrate the success of each approach.9,15 Overall, taking together the LE requirements for fragments to Indeed, a very recent review by Alex and Flocco tabulates 62 bind at different concentrations with the dramatic increase in fragment examples identified for 46 protein targets, illustrating availability of compounds as the upper limit of molecular size the rapidly gaining popularity of FBDD.64 However, for 29 of is increased, we can conclude that in a typical HCS campaign the 62 examples listed the fragments are potent enough to be the majority of the binding information will come only from detected by standard methods; here, we have focused on the largest compounds in the screening library. This is because examples where a FBDD technology is required to identify the the library is likely to be more highly populated with larger fragment hits.
compounds than smaller ones, and in any case only the larger 3.1. Evolving Fragments. Fragment binding can generally
compounds can be readily detected as hits for most targets.
be improved by substitution at one or more vectors with Figure 3 also illustrates that changes in LE are not linear additional functionality. This "fragment evolution" has proved with changes in HAC. This means that as fragments get smaller to be the most popular and effective approach to fragment for any given potency LE increases more quickly, and alterna- optimization. Structural information of the fragment hit (X-ray tively as drug sized ligands get larger, LE decreases more or NMR) is used to design larger molecules that pick up slowly. The result of this parabolic mathematical relationship additional protein-ligand interactions resulting in improved is that LE is very sensitive for small fragments (small changes affinity for the target. A requirement for success is that the in potency cause quite large changes in LE) but relatively fragment acts as an "anchor" and does not change its binding insensitive for larger compounds (LE does not change as mode during its evolution to a potent lead.
noticeably as HAC increases). Overall, therefore, it is important 3.1.1. -Secretase (BACE-1). The aspartyl protease enzyme
to use LE as a guide for ranking of fragment hits and also to -secretase is considered a very challenging target by the assess the success of optimization of fragments into potent larger industry, but the biological evidence is compelling that its ligands but not to overinterpret absolute LE values.
inhibition will be useful in the treatment of Alzheimer's disease.
In conclusion, it is our belief that FBDD is most powerful Mixtures of fragments have been screened using a 1D NMR when screening a library of smaller fragments (100-250 Da) screening approach, and fragments such as 1 (Scheme 1) were
using very sensitive biophysical methods at millimolar concen- identified with millimolar affinity as determined by surface trations. These smaller fragments have the best chance of finding plasmon resonance (SPR).65 Protein-ligand X-ray crystal- productive interactions with a given target protein, without lography determined that the fragments were bound through the containing a mismatch or steric clash that would compromise isocytosine motif to the catalytic aspartic acid residues in the binding,62 while still delivering sufficiently high LE to be active site. This type of charged bidentate interaction expressed detectable. When larger compounds are used and screened at a a newly discovered pharmacophore for this class of enzyme.66 lower concentration (such as 200 µM), the results might be By use of pharmacophore and structure-based drug design disappointing for all but the most tractable targets. This suggests approaches, potency was initially improved to the micromolar that FBDD may have significant advantages over HTS (and inhibitor 2, and then further optimization gave the lead molecule
perhaps HCS) when employed against more challenging targets, 3 with submicromolar affinity.67
such as proteases, phosphatases, and protein-protein interactions.
3.1.2. Urokinase (uPA). In a very recent example, the (R)-
enantiomer of the fragment-sized oral drug mexiletine 4 (Scheme
3. Optimization of Fragments
2) was identified as a hit binding to the active site of the serine The following section outlines a number of successful protease urokinase (urokinase-type plasminogen activator, uPA) strategies for the optimization of fragment hits to leads and in by X-ray crystallographic screening.68 Despite having potency one case to a clinical candidate. Many examples can now be too weak to be accurately measured, the fragment was optimized Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 Scheme 3. Fragment Evolution of PDE4 Inhibitors Using
chemistry that can take place at room temperature in an aqueous Scaffold-Based Drug Design environment in order to generate the inhibitors.
3.3.1. Carbonic Anhydrase. In this approach the protein
target of interest is used as a scaffold to bind adjacent fragments
containing suitable reactive functional groups. A chemical
reaction takes place, linking the two fragments to form an
inhibitor in situ in the active site of the protein. The example
shown (Scheme 6) used "click chemistry" to identify subna-
nomolar inhibitors of carbonic anhydrase, a zinc containing
metalloenzyme. Aromatic sulfonamides such as 15 were known
to strongly coordinate with the Zn2+ cation of the enzyme, and
the acetylene functional group was then appropriately positioned
for combination with complementary azides via a 1,3-dipolar
cycloaddition reaction that resulted in the triazole based
to a potent lead molecule 5, using structure-based design
inhibitors. For example, compound 16 showed a 185-fold
approaches. The lead compound was additionally shown to have improvement in binding affinity over the initial fragment.74 promising selectivity vs related proteases and a pharmacokinetic 3.4. Fragment Tethering. This strategy relies on the forma-
profile similar to mexiletine itself, with good oral bioavailability tion of a disulfide bond between a chemically reactive fragment and long half-life in rat. This example illustrates the value of and a cysteine residue in the target protein. Fragments with the screening low molecular weight drugs as part of a fragment greatest affinity for the protein within the vicinity of the cysteine form the most stable disulfide bonds and are readily detected 3.1.3. Phosphodiesterase 4 (PDE4). Researchers at Plexx-
by mass spectrometry.75 ikon have screened a library of around 20 000 scaffold-basedcompounds (molecular weight range 125-350 Da) against 5 3.4.1. Caspase-1. A nice example of this tethering approach
PDE family members using HCS. A follow-up screen using has been reported for screening and optimization against X-ray crystallography identified the pyrazole ester 6 (Scheme
caspase-1 (Scheme 7). First, a key recognition unit (the aspartyl 3). After just two rounds of chemical synthesis, a number of group of the substrate) was covalently linked to the active site low nanomolar potency compounds were discovered, including cysteine of caspase-1 to give adduct 17. Next, thiol fragments
7, equating to a 4000-fold increase in affinity.69
capable of forming a disulfide bond with this enzyme complex 3.2. Combining Fragments. In some cases more than one
were screened and hits such as 18 were identified by mass
fragment can be detected binding to nonoverlapping regions of spectrometry; these groups were later shown to bind into the a protein. It may then be productive to link such fragments, P4 pocket of the enzyme by X-ray crystallography. Replacement either directly or through a suitable linker. The challenge has of both the disulfide bond with a simple methylene linker and often been identifying a suitable linker strategy that allows both the covalent cysteine binder with a reversibly binding aldehyde fragments to maintain their original binding modes when resulted in compound 19 with submicromolar affinity.76 Further
combined in the new hybrid ligand and consequently retain optimization by rigidifying the linker unit produced a ligand acceptable LE.
efficient low nanomolar inhibitor 20.77
3.2.1. Bcl-X
3.5. Using Fragments as Substrates. Ellman's group have
Lead generation for the protein-protein anticancer target Bcl-X developed a novel fragment-based screening method called L was explored using a high-throughput NMR-based method "SAR by NMR".70 A chemical library of substrate activity screening (SAS) for the efficient development small molecules was screened for their potential to bind to the of novel nonpeptidic protease inhibitors.78,79 Here, the fragments large highly lipophilic BH-3 binding groove of Bcl-X being screened are substrates of the protease and are only later family member. In this way fragments 8 and 9 were found to
converted into inhibitors. This approach has also been recently bind to distinct but proximal subsites within the binding groove modified and employed for the discovery of novel protein of the active site (Scheme 4). By use of NMR derived structural tyrosine phosphatase inhibitors.
information and knowledge of key binding points for the native 3.5.1. Protein Tyrosine Phosphatase B (PtpB). A spectro-
substrate (BAK), the two fragments were linked via an alkene photometric method was employed to screen a diverse library functionality and then optimized for potency to give the low of O-aryl phosphates as potential substrates for the bacterial micromolar compound 10. Significant optimization efforts led
protein tyrosine phosphatase B.80 PtpB has potential as a target to the discovery of 11, an agent suitable for oral dosing, which
for the treatment of tuberculosis. Compound 21 was used as a
is currently in phase I/IIa cancer clinical trials (Table 2).
standard and had a KM of approximately 1 mM. The screen 3.2.2. Thrombin. A small targeted library of fragments
identified the biphenyl analogue 22 as a significantly better
(selected using virtual screening) was screened by soaking into substrate (Scheme 8). Systematic modification of the biaryl protein crystals of thrombin, and fragment hits were detected motif, followed by introduction of a suitable phosphate group directly by X-ray crystallography.71 A number of neutral S1 site isostere, resulted in compound 23, which is the most potent
binders were identified (Scheme 5), and one of these, fragment inhibitor of PtpB reported in the literature to date.
12, was linked to a larger ligand 13 which spanned the S2 and
3.6. Using Fragments To Identify New Binding Sites.
S4 subsites (derived from a virtual screening hit). With this Protein targets almost always have a "hot spot" within their fragment linking strategy, highly potent hybrid inhibitors such binding sites, for example, the hinge region within the ATP as 14 were discovered. The fragment binding mode of 12 is
binding pocket of a kinase enzyme or the metal ion in a matrix illustrated in Table 1, entry 9 (section 4.2.9).
metalloproteinase. It is often possible to block this hot spot by 3.3. In Situ Fragment Linking. Fragment linking has been
binding of one fragment and then probe the resulting protein-ligand taken a step further by the use of in situ approaches such as complex for a second binding event using a further set of dynamic combinatorial chemistry72 and in particular the use of fragments. In this way, fragment-based screening may be used "click chemistry".73 Here, there is a requirement for mild to identify molecules that bind to adjacent binding sites, Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 3667
Scheme 4. Fragment Linking for the Bcl-2 Family of Proteins Leading to the Identification of 11
Scheme 5. Fragment Linking Leading to Potent Non-Amidine
binding. With structural data derived from both NMR and X-ray Containing Inhibitors of the Blood Coagulation Target Thrombin experiments, a fragment-linking strategy was employed to design
more potent molecules such as 26 (K )
4 µM). This elegant example highlights the benefits of using multiple techniques for
generating structural data. The X-ray crystal structure suggested
a different binding mode for fragment 25 compared with the
NMR structure solution, and in this case it was the NMR data
that led to the design of the lead compound 26.
3.7. Computational Approaches. A variety of computational
tools have been developed or adapted to support the differentphases of fragment-based drug discovery programs, and this isa rapidly developing area. The first step in FBDD is to developlibraries of fragments that will be subsequently screened againstthe target. The principle considerations were touched on in Scheme 6. In Situ Click Chemistry Was Used To Identify
section 2.5, and most approaches have used chemoinformatics Subnanomolar Inhibitors of Carbonic Anhydrase, a Zinc and modeling tools as the starting point to generate fragment Containing Metalloenzyme libraries. The computational deconstruction of drugs into theirconstituent fragments is one of the most commonly usedmethods to build fragment libraries.82 Lepre and colleagues fromVertex Pharmaceuticals were one of the first groups to use thisapproach and generated a library containing less than 200fragments, which they called a SHAPES library, for NMRscreening.83 Similarly, a retrosynthetic combinatorial analysisprocedure (RECAP) has been used by Lewell et al. to identifyrecurring fragments from known drugs.84 Recently, Kolb andCaflisch have developed a computer program called DAIM(decomposition and identification of molecules), which is ableto disassemble molecules into mainly rigid fragments. DAIM,in conjunction with docking protocols, has been successfully potentially allowing for a fragment-linking approach to subse- applied to identify inhibitors of -secretase (BACE-1).85 quently be employed. In the case of more distant allosteric Computational methods are also being widely used to try to binding sites being identified, this also gives the option of identify potential hits from virtual fragment libraries. To date, modulating a protein's action via an alternative mechanism.
structure-based computational methods have been the most 3.6.1. Heat Shock Protein 90 (Hsp90). Huth et al. have
successful in this field. It can be difficult to generate good SAR identified low micromolar aminopyrimidine based inhibitors of for early fragment hits because they exhibit such low binding Hsp90 such as compound 24 (Scheme 9) using a fragment-
affinity that is then difficult to detect and compare reliably in a based NMR screening approach.81 Although this series of biological assay. As a result of this, successfully predicting the compounds could readily be optimized to give nanomolar binding mode of a fragment bound to its target, when this potency ligands, the lack of novelty of the leads was considered information is not available, is crucial in order to develop an issue. In an attempt to identify more novel inhibitors, a strategies to evolve a hit into a more potent lead molecule.
fragment library was screened against Hsp90 using a 2D NMR Molecular docking has been successfully used for a number of method in the presence of compound 24. In this way, fragment
years to predict the binding modes of druglike compounds, and 25 was identified that had K )
150 µM in the presence of 24
more recently, this methodology has been applied to dock but K > 5000 µM in the absence of 24, suggesting cooperative
fragment-like molecules.86 A typical work flow in FBDD Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 Scheme 7. Identification of Caspase-1 Inhibitors Using a Fragment Tethering Approach
Scheme 8. Identification of a Potent Inhibitor of PtpB through Fragment-Based Substrate Activity Screening
Scheme 9. Identification of Second Site Binders in HSP90 by
are often dependent on ligand size; therefore, virtual fragment libraries should preferably contain molecules of similar size andnumbers of functional groups in order to demonstrate goodenrichment in a virtual screening experiment.94 Despite theseissues, the success of biophysical screening in FBDD hasencouraged the modeling community to try to develop new andmore reliable tools with the aim of improving the modeling offragment binding for virtual screening. Better treatment ofsolvation and electrostatics85 and the use of grand canonicalMonte Carlo methods to calculate binding free energies95 haveall shown promise and offer new directions in fragmentmodeling. Two recent examples that illustrate how computa- consists of docking a fragment into the binding site of interest, tional methods have been used for FBDD and in which new choosing the best orientation and then using this as a starting potent inhibitors have been discovered are given below.
point for the attachment of substituents with the aim of targeting 3.7.1. Dipeptidyl Peptidase IV (DPP-IV). Rummey et al.
a new area within the binding site where supplementary used an in silico docking approach to identify fragments that interactions might be made. A number of SBDD tools have been bound within the S1 pocket of DPP-IV.96 A fragment database developed with this work flow in mind, and these methods have of approximately 10 000 small primary aliphatic amines was been recently reviewed.59,87 There are some excellent examples assembled and docked using FlexX. A range of -phenethy- in the literature in which docking approaches have identified lamine based compounds were identified, and fragment 27
virtual fragment hits that have helped to develop inhibitors for (Scheme 10) was chosen for optimization. X-ray crystallography a range of different targets that include thrombin,88 factor Xa,89 confirmed that the phenyl ring of the fragment was located in cathepsin D,90 TGT,91 and CDK4.92 However, the low molecularweight and low complexity of fragments have highlighted some the hydrophobic S1 pocket that accommodates a proline in the limitations in docking methodology. Often, scoring functions substrate. The amino group was shown to form multiple are not able to energetically distinguish native from irrelevant hydrogen bonds to three acidic residues. When this was used poses,93 and pharmacophoric constraints need to be applied to as a start point, structure-based optimization resulted in the improve results.94 Moreover, most of the scoring functions used discovery of a series of potent DPP-IV inhibitors, such as 28.97
in molecular docking contain a number of crude approximations Another example of the fragment-based discovery of DPP-IV of the factors involved in binding and have been optimized for inhibitors is given in Table 1, entry 4, which shows the crystal druglike ligands. Not surprisingly then, they do not perform structure of a substrate-based fragment bound to DPP-IV. Its well with fragment-like molecules. Finally, scoring functions subsequent optimization is illustrated in part ii of Figure 4.
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 3669
Scheme 10. Identification of Potent DPP-IV Inhibitors Using in
the key recognition features required for optimization to this Silico Fragment-Based Screening larger inhibitor,103 another that the starting inhibitor was notoptimal.14 A third possibility is that in some cases it may notbe possible to break a potent inhibitor down into very smallfragments because synergy between two nonadjacent bindinginteractions is required for efficient binding. If this is the case,it would follow that one might not expect deconstruction of apotent lead to always derive high quality fragments. In contrast,it has been our observation at Astex that, when starting fromefficient fragment hits, a fragment both maintains its interactionsand can be used to readily derive a potent lead during itsoptimization. In the next section we look in more detail at how 3.7.2. Aurora A Kinase. Warner et al. screened a virtual
representative fragments bind to their protein targets and we fragment library of ∼70 000 compounds (including a limited consider how a number of these fragments compare with larger, number of known kinase inhibitor fragments) against a 3D more optimized molecules derived from them. In doing so, this homology model of aurora A kinase.98 Subsequent structure- question of how reliably fragments can be optimized with guided optimization was then carried out using a published maintenance of their original binding mode will be further crystal structure (1MQ4) from the PDB. The known kinase scaffold 29 (Scheme 11) was identified in silico using the scoring
function LUDI and was then confirmed to have inhibitory
4. Fragment-Protein Complexes
activity against aurora A (IC 3 µM). Similarly, compound 30 was proposed to bind in the phosphate binding pocket (IC
In this section we review 12 representative fragment-protein ) 21 µM). A fragment linking strategy was used to give complexes available in the PDB. Here, we have attempted to compound 31, a nanomolar potency inhibitor and a useful
illustrate the types of interactions typically observed between starting point for further medicinal chemistry optimization.
fragments and their target proteins and have selected examples However, this example highlights the difficulty of linking two that highlight some of the issues and opportunities with FBDD.
relatively ligand efficient fragments together, as the resulting 4.1. Selection of Representative Fragment-Protein Com-
larger molecule has a significantly lower LE and high MWT.
plexes. The Protein Data Bank is an excellent source of
3.8. Deconstructing Leads into Fragments. Taking a lead
information about how ligands interact with proteins (www.
molecule, breaking it into key fragments, and identifying where pdb.org). Various groups have used this data to implement, test, they bind can provide a valuable source of structural informa- and improve computational tools for structure-based drug tion. At Astex, we have routinely applied this approach to derive design.14,104–108 Recently, Hartshorn et al. have published a focused fragments for targets using pre-existing lead series from procedure to extract, analyze, and classify protein-ligand X-ray the medicinal chemistry literature. A number of other groups crystal structures from the PDB.109 This process delivered a have also explored the idea of fragmenting known drug-sized set of clusters of protein-ligand complexes, where each cluster ligands.28,53,99–101 In particular, Hajduk has reported an analysis included only high-quality PDB structures (high resolution and of the deconstruction of 18 highly optimized inhibitors into optimal electron density around the ligand). In addition, most successively smaller component compounds and fragments.53 of the selected complexes were associated with targets of Perhaps unexpectedly, the fragments and final compounds were relevance to drug discovery or to agrochemicals and contained found to have similar LEs, suggesting that it is possible to only ligands that obeyed Lipinski's rules. This high-quality maintain the LE of the starting ligand during the optimization subset of the PDB (HQPDB) offers an ideal starting point to of fragments. In fact, at Astex our experience is consistent with review public domain structures that contain fragment-like these observations, and any fragment considered for optimization ligands. To the best of our knowledge, this type of analysis has will have a minimum LE of 0.3 in order that an optimized high not been done before.
potency lead will have a molecular weight of <500 Da. This A simple way to extract relevant fragment-containing struc- criterion would only be relaxed for very challenging targets with tures from this HQPDB subset was to apply a ligand MWT borderline druggability such as protein-protein interactions, cutoff of 300 Da and to exclude ligands that did not contain at targets that are now increasingly being considered for FBDD least one aliphatic or aromatic ring. After these filters were approaches.102 In our experience it is not reasonable to expect applied to the 85 clusters in the HQPDB, only 45 of them LE to increase significantly during the fragment optimization remained populated. All populated clusters were next inspected process unless the LE of the starting fragment can be im- using a graphical window containing AstexViewer in order to mediately improved by optimization of its binding mode without remove solvent-like molecules, fragments not bound within the increasing (or even by reducing) its size.
active site of the protein in question, and other "false-positive" Babaoglu et al. have deconstructed a 1 µM potency -lac- fragments originated by connectivity errors in the PDB file.110 tamase inhibitor into four small, low affinity fragments.99 They Finally, 12 diverse representative protein-ligand structures were obtained experimental binding modes for all the fragments (X- selected for the scope of this review. In particular, an effort ray crystal structures), but only one bound in the manner has been made to represent a broad range of proteins and also predicted from the original inhibitor. One fragment bound in a to select examples that illustrate the range of interactions seen new pocket, two via a novel binding mode within the existing in the broader context of the public domain fragment-ligand pocket, and only the molecule with slightly higher complexity complexes that were inspected.
was found to bind in the same manner as the original compound.
4.2. Description of 12 Representative Fragment-Protein
This report has been much debated by those engaged in FBDD Complexes. Summarized in Table 1 is an illustration and the
and has cast some doubts on how reliable the binding mode of associated data for each fragment-protein complex being a fragment will be during its subsequent optimization. One considered. Each entry contains the 2D structure of the ligand possible explanation is that only one of the fragments contained with associated MWT, binding affinity, and LE, a 3D repre- Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 Scheme 11. Virtual Fragment Library Screening against Aurora A Kinase
sentation of the ligand in the binding site, and a detailed 4.2.4. Dipeptidyl Peptidase IV (DPP-IV). Entry 4 describes
description of the binding mode. The entries are ordered by a very efficient substrate based fragment bound to DPP-IV.116 submission date of the X-ray crystal structure in the PDB This fragment 34 has been evolved into larger and more potent
database (only structures deposited after August 11, 2000, have inhibitors such as 35 (Figure 4ii) (a related example was also
been included). Additionally, Figure 4 contains schemes for given in section 3.7.1).117 X-ray crystal structures show that examples where potent inhibitors in the same chemical series advanced compounds such as compound 35 (PDB code 2FJP)
have been reported; in each case the advanced lead is colored conserve the original binding mode of the fragment.
according to the moiety that originated from the starting 4.2.5. Neuronal Nitric Oxide Synthase (n-NOS). Entry 5
fragment. In some cases the leads were developed from the shows a fragment bound to the catalytic site in the oxygenase inhibitors using FBDD, but in others the fragment was identified domain of n-NOS.118,119 This relatively small pocket is an later by lead deconstruction. Here, we also describe some of example of a good binding site for fragment-like molecules our observations for each of the examples, and a number of because it has a volume similar to that of the fragment. To date, more general conclusions are discussed in section 5 of this the PDB contains 17 NOS structures with nonpeptidic fragment- like molecules bound to the catalytic site in the oxygenase 4.2.1. Cyclooxygenase-1 (COX-1). Entry 1 contains the
domain. All of these fragments interact with the heme group structure of the active enantiomer of ibuprofen complexed with through the formation of stacking interactions and/or by COX-1. Ibuprofen is an example of a marketed oral drug having chelation with the iron atom within the heme itself. The fragment fragment-like properties. The carboxylate group plays a key role in entry 5 is interesting not only because it binds to this small in the recognition of the molecule within the COX-1 binding pocket but also because it offers good vectors to target the area site, while the shape and the size of the lipophilic group are outside the catalytic site region, where it is known it is possible thought to be important in allowing the molecule to pass through to modulate the selectivity profile of the molecule for the various the cyclooxygenase channel itself rather than being critical for NOS isoforms. When there are a number of fragments available affinity.111 This example illustrates the value of using historical that bind in a similar manner, choosing examples that allow drug molecules and also simplified scaffolds derived from ready access synthetically to other regions of the active site of known drugs for fragment screening.
interest is a useful strategy for fragment evolution.
4.2.2. Urokinase (uPA). Entry 2 shows an example of an
4.2.6. Cyclin Dependent Kinase 2 (CDK2). Entry 6 shows
aminobenzimidazole fragment bound to urokinase. Hajduk et 2-amino-6-chloropyrazine 36 bound to the hinge region of
al. used NMR-based screening to identify this novel class of CDK2.37,120 The fragment picks up interactions similar to those urokinase inhibitors from a library of more than 3000 of the native ligand ATP. The compound has low affinity (IC50 compounds.112,113 X-ray crystallography was subsequently used ) 350 µM), but because of its low MWT, it has good ligand to confirm the binding mode of the 2-aminobenzimidazole hit.
efficiency (LE ) 0.59). Substitution of the aminopyrazine 36
This is one of the earliest reported examples of FBDD where a with an arylsulfonamide gave only a modest jump in activity crystal structure of the fragment has been deposited in the PDB.
to compound 37 (IC
9.1 µM) and a reduction in LE (0.38), The fragment forms a number of H-bond interactions deep even if X-ray crystallography confirmed (unpublished data) that within the S1 pocket of the protease. The fragments in this entry compound 37 retained the binding mode observed in the original
and in entry 1 are both charged and likely to have high aqueous fragment 36 (Figure 4iii). This is hard to rationalize, as good
solubility that would facilitate fragment screening.
hydrophobic contacts are made between the aryl group and 4.2.3. Estrogen Receptor r (ER). The ligand in entry 3 of
hydrophobic residues at the protein surface and the binding of the table is a fragment 32 of the ER antagonist raloxifene, 33
the pyrazine portion of the molecule is identical to that of the (Figure 4i).114 The benzothiophene is a rigid scaffold of a starting fragment. An additional H-bonding interaction between suitable length to target two distal polar regions of the binding the NH2 of the sulfonamide and Asp86 is observed, but this is site with hydroxyl groups. It also offers good growing points at the expense of breaking an intramolecular salt bridge between to other parts of the active site. The raloxifene structure (not Asp86 and Lys89. This latter point may explain the modest jump shown; PDB code 1ERR) indicates that the introduction of the in activity. Optimization of this series at Astex was subsequently large substituent at the 3-position of the benzothiophene altered halted in favor of other fragments in which LE was more easily the binding mode compared with the fragment 32 but that the
maintained during optimization and which eventually led to two phenol-binding "hot spot" within the active site is engaged in clinical candidates (Table 2), illustrating that it is important not the same manner.115 These data suggest that smaller, low to focus merely on increasing potency but instead on maintaining potency, phenol based fragments would be expected to bind to LE for successful fragment optimization.
the receptor if raloxifene were to be deconstructed, as outlined 4.2.7. Phenylethanolamine N-Methyl Transferase (PNMT).
is section 3.8.
Entry 7 shows the X-ray crystal structure of a very active and Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 3671
Table 1. Representative Fragment-Protein Complexes Selected from the PDB
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 Table 1 (Continued)
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 3673
Table 1 (Continued)
efficient tetrahydroisoquinoline (THIQ) based inhibitor 38 bound
of epinephrine. Grunewald et al. used this fragment as a lead to the active site of PNMT, a key enzyme in the biosynthesis compound to design in selectivity over the R2-adrenoceptor and Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 Figure 4. Examples of fragment-sized molecules available in the PDB and their associated drug-sized leads. In some cases the fragment information
was used directly to develop the lead, but in other cases the fragment information was only available retrospectively and the lead was developed
by standard approaches (see text): (i) ER, (ii) DPP-IV, (iii) CDK2, (iv) PNMT, (v) TGT, (vi) BACE-1.
modulate the physiochemical properties of the series to improve forms six hydrogen bonds in addition to good lipophilic CNS penetration (Figure 4iv).121 Compound 39 had good affinity
interactions with the enzyme, resulting in a highly ligand for the target enzyme and was found to be 19000-fold selective efficient compound.124 The aminoquinazoline has been devel- over the R2-adrenoceptor. The X-ray crystal structure for 39 is
oped into larger leadlike molecules using SBDD.125 Figure 4v not available in the PDB; however, the structures of closely shows an aminoquinazoline derivative 41 that retains the binding
related analogues (e.g., PDB codes 2G8N and 2G71) retained mode of the original fragment 40 (compare PDB entries 1S39
the binding mode of the original THIQ core. This example and 1Y57). However, the side chains of 41 are not used to
illustrates how having high LE leaves scope to affect the increase the binding affinity but instead to modulate the enzyme druglike properties of a molecule, giving "molecular weight head kinetic profile of the series; the original fragment 40 is a
room" to facilitate the lead optimization process for a given noncompetitive TGT inhibitor, whereas the evolved molecule series of inhibitors.
41 is a competitive TGT inhibitor. This is another example of
4.2.8. Methionine Aminopeptidase 2 (MetAp2). Entry 8
how good LE can leave room for optimization of other desirable is an example of a fragment acting as a metal chelator.122 The properties, as well as high potency.
strong interaction between the 1,2,3-triazole and the two cobalt 4.2.11. -Secretase (BACE-1). Entry 11 shows an aminoiso-
cations confers a high degree of LE to the fragment. Considering quinoline fragment bound to the two catalytic aspartic acid its small size and high LE, aryltriazole based inhibitors of this residues of -secretase (BACE-1). This type of charged bidentate type may provide a promising template for the design of new interaction represents a new pharmacophore for this class of inhibitors for methionine aminopeptidase, a target relevant in enzymes.66 With this information, 42 was developed into
oncology. Indeed, related triazoles, which show similar interac- alternative fragments such as compound 43 using virtual
tions with the active site cobalt ions, have recently been screening by docking using an aminopyridine/amidine pharma- disclosed.123 Metalloproteinases are another class of enzymes cophore derived from the structure 2OHK. SBDD was then for which FBDD is ideally suited because of the high LE applied to design inhibitor 44 that accessed both the S1 and S3
available from interacting with metal ions important in enzyme pockets of the protease (Figure 4vi).126 Compound 44 conserved
the binding mode observed in fragment 43 (see PDB entry
4.2.9. Thrombin. Entry 9 shows a fragment bound to the
2OHU). It is interesting to note that the relative position S1 pocket of thrombin. It is interesting to note that the ligand of Asp32 and Asp228 in -secretase resemble the positions of has good LE, even if binding is driven predominantly by Asp156 and Asp102 in TGT (entry 10). The binding mode of hydrophobic interactions.71 The optimization of this fragment isocytosine fragments discovered for -secretase are described via a linking approach has been described in section 3.2.2. This in section 3.1.1 and are quite reminiscent of the interactions example indicates that fragments need not have multiple polar made by the aminoquinazolinone fragment in TGT. This is a interactions to have good LE, especially if the target's active good example of how fragments have the potential to form site has deep and well defined pockets on its surface.
similar interactions in different and unrelated targets.
4.2.10. tRNA-Guanine Transglycosylase (TGT). Entry 10
4.2.12. Heat Shock Protein 90 (Hsp90). Entry 12 shows
shows a 2-aminoquinazolinone fragment bound to the catalytic 2-aminopyrimidine bound in the ATPase domain of HSP90. This site of TGT. In contrast to the previous example, the fragment structure is particularly noteworthy because the ligand makes Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 3675
only one hydrogen bond directly with the protein and three identify further positive interactions between the ligand and additional hydrogen bonds through conserved water molecules.
protein. Inhibitors of many tractable targets, such as kinases The fragment has undetectable binding affinity because it is so with their well defined ATP binding sites, have been discovered small. Despite this, however, this fragment can still be optimized by this approach. However, far more challenging targets, for and larger analogues with submicromolar potency and good LE example, the aspartyl protease BACE-1 for the potential have been reported in which the aminopyrimidine fragments treatment of Alzheimer's disease, have also been successfully form identical interactions with the protein (see section 3.6.1).81 targeted (section 3.1.1). In this case many groups had failed to This illustrates how even very small fragments serve as useful find tractable nonpeptidic lead series using traditional HTS, hits and can provide valuable binding information.
whereas FBDD proved to be a very effective alternative. Where 4.3. Summary. In this section we have used HQPDB to
more than one fragment has been observed to bind to different obtain a selection of high quality protein structures that contain regions of a target protein, a fragment-linking strategy can be fragment-like ligands. Twelve representative examples were applied. This methodology has been used to identify very high selected in order to highlight some general aspects and properties affinity lead molecules, for example, by Abbott to design a Bcl- of fragments. Below we report a short summary of our findings: XL inhibitor now in clinical trials. However, very precise linking (1) The binding mode of fragments is often driven by good of the two fragments must be achieved in each case to give the quality polar interactions and also by good shape matching with expected benefits of superadditivity, and so far, the LEs of the binding site. However, fragments can sometimes also bind inhibitors have generally been lower than those of the starting purely through hydrophobic contacts and still have good LE.
fragments.54 We have seen many other innovative developments (2) Fragments can generally be optimized into potent inhibi- over the past 4 years that have helped this area progress rapidly, tors while maintaining the original binding mode of the starting including novel approaches to HCS such as substrate activity screening, further applications of fragment tethering, improve- (3) Fragments have the potential to be more promiscuous ments in virtual screening, and the harnessing of the chemo- binders than larger molecules.
specific reactivity of fragments using in situ click chemistry.
(4) Ligand efficient fragments can be evolved into relatively Section 4 illustrates, for those remaining doubters of the value small lead molecules (with high LE), leaving a "molecular of FBDD, that in fact fragments have been with us for many weight window" to allow for optimization of other properties years (but have just not been recognized as such) and also that such us PK or selectivity (with some consequent reduction in they do not differ in the types of interactions they form with their targets compared to larger, more potent ligands (exceptthat they generally form fewer of them). Indeed, we can see 5. Discussion and Conclusions
from the binding of ibuprofen (entry 1 in Table 1) that thecompound looks grossly similar in the number of interactions In this review we have attempted to examine the trends and and van der Waals contacts as many of the other examples given, developments in FBDD and fragment screening over the past 4years. To date, fragment screening has been applied in three and this compound is a licensed drug. An analysis of our in- main ways. First, its use as a primary screening tool has been house oral drug database containing 1195 licensed oral drugs well documented with many examples being reported, some (an updated version of the database described by Vieth et al.) being outlined earlier. This approach is popular within academia indicates that 211 drugs (17.7%) obey the "rule of 3" discussed and biotechnology companies. Second, FBDD is being used earlier, highlighting how small, simple molecules have often successfully to identify hits when a HTS campaign has failed been a rich source of drugs historically.82 However, given the to yield useful results against difficult targets.67 Third, it is challenging nature of many modern drug targets, it may be starting to become more common within large pharmaceutical unrealistic to expect this proportion to be maintained for drugs companies to run FBDD in parallel with HTS campaigns.12 This in the future. It can also be seen from an examination of the last trend is due to the undoubted success of the methodology fragment-protein complexes in the table and consideration of for many targets. However, there is still skepticism within the their respective affinities that it is not easy to predict with any pharmaceutical industry that the optimization process to give confidence the potency of a fragment by considering how it is nanomolar potency leads from very low affinity hits can be more interacting with its target. This is also true of drug-sized ligands broadly and routinely achieved. The examples discussed herein where nanomolar or micromolar compounds often look very and in other reviews indicate that fragment-based approaches similar in their binding modes within any given target protein.
have been successful for a wide range of targets, and our Potency (and therefore LE) is not simply the sum of the number experience here at Astex is that if FBDD identifies a number of interactions and hydrophobic contacts a molecule makes; it of hits with good LE (>0.3), then optimization of at least one is far more complex than that. This is why it is so hard to rank of these series to potent and efficient low nanomolar potency molecules in silico by virtual screening with any confidence leads will generally follow, provided that there is reliable access and particularly difficult for fragments. What we can conclude, to protein-ligand structural information.
though, is that the maximal LE achievable will be a function The selected examples in section 3 illustrate both the broad of the binding site (not the fragment) and that for many targets applicability of FBDD and the rate of new developments in the very high LE has been observed historically, allowing easy underlying fragment binding detection technology. Fragment detection of fragment binding and subsequent progression to evolution has been by far the most successful method of low MWT drugs. Entry 7 in Table 1 illustrates this very clearly fragment optimization. It is conceptually the most straightfor- (PNMT). Here, the fragment-sized compound has extraordinary ward and particularly when allied with a high degree of potency and LE (IC 0.003 µM, LE ) 0.97) but has only a structural information (e.g., from NMR or X-ray crystal- single polar interaction with the protein and quite limited van lography) gives the medicinal chemist a valuable advantage in der Waals contact. The high LE observed is as much a function the validation and subsequent optimization of a hit. The fragment of the binding site as the fragment itself. This type of reasoning itself binds to the "hot spot" in the active site, with structural has led Hajduk et al. to propose that fragment screening can be information being used to guide iterative cycles of design to used as an approach to examine target druggability, whereby if Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 no hits are observed from FBDD, then the target can be Table 2. Clinical and Preclinical Candidates Derived from Fragments
considered intractable to small molecules.127 The recent advent of FBDD is then simply a broadening of the use of fragments to targets that have a more typical maximal LE rather than limiting the use of fragments to highly tractable targets. For example, if we compare entries 10 and 11, we can see the fragments bind through a similar network of hydrogen bonds and they are of similar size. In the first case (tRNA-guanine transglycosylase) the compound has a potency of IC SGX Pharmaceuticals µM (LE ) 0.65), while in the latter system (BACE-1) the 50 ≈ 2000 µM (LE ≈ 0.33). In the first example the compound was not considered to be a fragment, but instead a lead molecule, because of its high potency. In the latter case, FBDD methods were required for hit finding (i.e., a highly sensitive detection system (in this case X-ray crystallography)) and careful subsequent optimization was needed using SBDD SGX Pharmaceuticals approaches to convert the fragment into a lead series.126 We can also see from examination of the fragment binding modes ethyl]-1H-indole-6-carboxamide. b Structure not disclosed. c Scheme 4, 11.
d (1S)-1-((4S)-2,2-Dimethyl-1,3-dioxolan-4-yl)-2-[[4-[4′-(trifluoromethoxy)-
in the two proteins (entries 10 and 11) that we might anticipate that both of the fragments could bind to each of these proteins.
This binding promiscuity is desirable within a fragment library lic acid ethylamide. f Propane-1-sulfonic acid [3-(5-chloro-1H-pyrrolo- and relates back to the sampling of chemical diversity discussed in section 2.5. Examination of the elaborated inhibitors in parts v and vi of Figure 4 together with a consideration of the shapeof each of the active sites suggests that the larger lead compound place in the armory of tools pharmaceutical companies employ would not be cross-reactive, illustrating how binding information to prosecute new targets. More than the underlying technology will often be lost as MWT and chemical complexity increase.
itself, the legacy of FBDD will be conceptual and that "less is Finally, in considering each of the eight examples where more more"; breaking down the problem of finding a drug molecule potent lead molecules (Figure 4, section 3.2.2, and section 3.6.1) into its smallest component parts opens up new opportunities have been developed from the fragments pictured in Table 1, to solve challenging problems. The next chapter in the develop- we see that the whole fragment or in one case (Figure 4i) the ment of the field is likely to be the evaluation of FBDD against key region of the molecule maintains binding to the "hot spot" targets where high resolution structural information is more of each active site. In addition, in many of the examples difficult to obtain. Only in this way can these key concepts be described in section 3, there are structural data that support that applied to the whole gamut of drug targets that the industry the starting fragment has been evolved to a lead compound in wants to work on. There are already companies committed to which the core derived from the fragment recapitulates the this approach. AstraZeneca have developed a work flow to tackle interactions seen in original hit. Along with our own observa- G-protein-coupled receptors and routinely screens a 20 000 tions at Astex, this further supports that fragment evolution is membered fragment library against all its targets, independent generally a reliable process, despite the observation that of the likelihood of obtaining protein-ligand structures.129 fragment deconstruction may not always be so.99 Ultimately, however, for success where experimental structural FBDD cannot have truly been said to have come of age until data are limited, methodology will need to be greatly improved a drug is launched that is derived from a low potency fragment.
to allow detailed SAR information to be generated in the 100 Table 2 lists the clinical and preclinical candidates and programs µM to 1 mM potency range so that empirical medicinal for which there is a clear statement in the public domain that chemistry optimization becomes possible for fragments. This, the candidate drug has been derived using fragment-based drug coupled with continued breakthroughs in structural biology, discovery. In addition, Hajduk and Greer give a more compre- particularly improving our understanding of the active sites of hensive table of lead molecules derived from fragments.16 receptors (rather than enzymes) and further developments in Perhaps we would not expect a drug to have been born out of modeling and computational chemistry, may soon make starting FBDD quite yet, as it took many years for SBDD to claim its a drug discovery project from a fragment part of standard first successful drug discovery projects.128 However, these data medicinal chemistry best practice rather than the specialized support that the screening and optimization of fragments are technology it once was.130 beginning to have an impact on the clinical pipeline of Acknowledgment. The authors thank Sahil Patel for prepar-
companies. Of particular note in Table 2 is the discovery of 11
ing and submitting the PDB structure for Table 1, entry 12, targeting the protein-protein interaction Bcl-XL (section 3.2.1) Paul Mortenson for generating the data used to produce Figure currently in phase 1/2a clinical trials for cancer. This is one of 2, Charlotte Cartwright for preparation of Figures 2 and 3, the first non-natural product clinical agents developed against Martyn Frederickson, Brian Williams, Emma Vickerstaffe, and any protein-protein interaction drug target, and in a recent Charlotte Griffiths-Jones for proofreading, and David Rees, review on the topic, Wells and McClendon suggested that Chris Murray, and Harren Jhoti for useful discussions.
fragment screening may be a more fruitful approach than HTSfor this extremely challenging area.102 So what does the future hold? One thing is certain. FBDD Miles Congreve is Director of Chemistry at Astex Therapeutics
will not be a panacea that will solve all of our problems. Like in Cambridge, U.K., where he is responsible for the company's structure-based drug design, computational chemistry, HTS, and fragment screening collection and for its hits to leads chemistry combinatorial chemistry before it, FBDD will find an appropriate function. He joined Astex in 2001. He previously held various Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 3677
positions at GlaxoSmithKline (1993-2001), working on a broad strategic advances and lessons learned. Nat. ReV. Drug DiscoVery range of medicinal chemistry and hits to leads projects at Ware 2007, 6, 211–219.
and then Stevenage before moving to Cambridge in 1999 to run (17) Jhoti, H.; Cleasby, A.; Verdonk, M.; Williams, G. Fragment-based screening using X-ray crystallography and NMR spectroscopy. Curr. the GlaxoWellcome chemistry research facility embedded in the Opin. Chem. Biol. 2007, 11, 485–493.
University of Cambridge chemistry department. His Ph.D. studies (18) Schade, M. Fragment-Based Lead Discovery by NMR. Frontiers in were also carried out at the University of Cambridge (1990-1993), Drug Design and DiscoVery; Bentham Science Publisher, Ltd.: Berlin, supervised by Professor Andrew Holmes.
Germany, 2007; pp 105-119.
Gianni Chessari is a Senior Computational Chemist at Astex
(19) Schuffenhauer, A.; Ruedisser, S.; Marzinzik, A. L.; Jahnke, W.; Therapeutics, joining the company in 2002. He has been involved Blommers, M.; Selzer, P.; Jacoby, E. Library design for fragment in the construction of the fragment corporate libraries and also based screening. Curr. Top. Med. Chem. 2005, 5, 751–762.
(20) Siegel, M. G.; Vieth, M. Drugs in other drugs: a new look at drugs contributed to the design of inhibitors for a number of drug targets as fragments. Drug DiscoVery Today 2007, 12, 71–79.
including thrombin, -secretase, urokinase, and Hsp90. Before (21) Wyss, D. F.; Eaton, H. L. Fragment-Based Approaches to Lead working at Astex, he carried out postdoctoral research at the Institute Discovery. Frontiers in Drug Design and DiscoVery; Bentham of Cancer Research in London (U.K.) and the CNRS in Lille Science Publisher, Ltd.: Berlin, Germany, 2007; pp 171-202.
(France). He received his Ph.D. in Organic Chemistry at the (22) Jahnke, W.; Erlanson, D. A. Fragment-Based Approaches in Drug University of Sheffield (U.K.) in 2000 under the supervision of DiscoVery; Wiley: Weinheim, Germany, 2007.
(23) Jhoti, H.; Leach, A. R. Structure Based Drug DiscoVery; Springer: Professor Christopher Hunter.
Dordrecht, The Netherlands, 2007.
Dominic Tisi is a Senior Structural Biologist working in the
(24) Shuker, S. B.; Hajduk, P. J.; Meadows, R. P.; Fesik, S. W.
Protein Structure group at Astex Therapeutics. Joining the company Discovering high-affinity ligands for proteins: SAR by NMR. Science in 2000, he has worked on several drug discovery projects including 1996, 274, 1531–1534.
JAK2, Aurora kinase, and PTP1B, performing fragment screens (25) Verlinde, C. L. M. J.; Kim, H.; Bernstein, B. E.; Mande, S. C.; Hol, using X-ray crystallography. He received his Ph.D. in Protein W. G. Antitrypanosomiasis Drug Development Based on Structures Crystallography from Birkbeck College, London, and GlaxoW- of Glycolytic Enzymes. Structure-Based Drug Design; MarcelDekker: New York, 1997; pp 365-394.
ellcome, Stevenage, in 1998 before carrying out postdoctoral (26) Vanwetswinkel, S.; Heetebrij, R. J.; van Duynhoven, J.; Hollander, research at Imperial College, London, prior to joining Astex J. G.; Filippov, D. V.; Hajduk, P. J.; Siegal, G. TINS, target immobilized NMR screening: an efficient and sensitive method for Andrew J. Woodhead is a Senior Chemist at Astex Therapeu-
ligand discovery. Chem. Biol. 2005, 12, 207–216.
tics. He has been with Astex since 2001 and has experience in (27) Bosch, J.; Robien, M. A.; Mehlin, C.; Boni, E.; Riechers, A.; Buckner, fragment hit identification and structure based drug design for a F. S.; Van Voorhis, W. C.; Myler, P. J.; Worthey, E. A.; DeTitta, number of targets including the oncology targets CDK2 and HSP90.
G.; Luft, J. R.; Lauricella, A.; Gulde, S.; Anderson, L. A.; Kalyuzhniy,O.; Neely, H. M.; Ross, J.; Earnest, T. N.; Soltis, M.; Schoenfeld, He began his medicinal chemistry career with Wyeth in 1992, L.; Zucker, F.; Merritt, E. A.; Fan, E.; Verlinde, C. L.; Hol, W. G.
working on CNS disorders such as migraine and depression. In Using fragment cocktail crystallography to assist inhibitor design of 1994 he joined Roche to focus initially on inflammatory diseases Trypanosoma brucei nucleoside 2-deoxyribosyltransferase. J. Med. and then moved to the virology department before finally joining Chem. 2006, 49, 5939–5946.
the lead generation chemistry group. He carried out his academic (28) Ciulli, A.; Williams, G.; Smith, A. G.; Blundell, T. L.; Abell, C.
studies at De Montfort University, Leicester, U.K.
Probing hot spots at protein-ligand binding sites: a fragment-based
approach using biophysical methods. J. Med. Chem. 2006, 49, 4992–
(29) Betz, M.; Vogtherr, M.; Schieborr, U.; Elshorst, B.; Grimme, S.; Pescatore, B.; Langer, T.; Saxena, K.; Schwalbe, H. Chemical Biology (1) Fink, T.; Bruggesser, H.; Reymond, J. L. Virtual exploration of the of Kinases Studied by NMR Spectroscopy. Chemical Biology: From small-molecule chemical universe below 160 daltons. Angew. Chem., Small Molecules to Systems Biology and Drug Design; Wiley-VCH Int. Ed. 2005, 44, 1504–1508.
Verlag GmbH & Co. KGaA: Weinheim, Germany, 2007; pp 852- (2) Bohacek, R. S.; McMartin, C.; Guida, W. C. The art and practice of structure-based drug design: a molecular modeling perspective. Med. (30) Bretonnet, A. S.; Jochum, A.; Walker, O.; Krimm, I.; Goekjian, P.; Res. ReV. 1996, 16, 3–50.
Marcillat, O.; Lancelin, J. M. NMR screening applied to the fragment- (3) Erlanson, D. A.; McDowell, R. S.; O'Brien, T. Fragment-based drug based generation of inhibitors of creatine kinase exploiting a new discovery. J. Med. Chem. 2004, 47, 3463–3482.
interaction proximate to the ATP binding site. J. Med. Chem. 2007,
(4) Fattori, D. Molecular recognition: the fragment approach in lead generation. Drug DiscoVery Today 2004, 9, 229–238.
(31) Chen, J.; Zhang, Z.; Stebbins, J. L.; Zhang, X.; Hoffman, R.; Moore, (5) Rees, D. C.; Congreve, M.; Murray, C. W.; Carr, R. Fragment-based A.; Pellecchia, M. A fragment-based approach for the discovery of lead discovery. Nat. ReV. Drug DiscoVery 2004, 3, 660–672.
isoform-specific p38-alpha inhibitors. ACS Chem. Biol. 2007, 2, 329–
(6) Lundqvist, T. The devil is still in the detailssdriving early drug discovery forward with biophysical experimental methods. Curr. (32) Chung, F.; Tisne, C.; Lecourt, T.; Dardel, F.; Micouin, L. NMR- Opin. Drug DiscoVery DeV. 2005, 8, 513–519.
guided fragment-based approach for the design of tRNA(Lys3) (7) Moore, W. R., Jr. Maximizing discovery efficiency with a compu- ligands. Angew. Chem., Int. Ed. 2007, 46, 4489–4491.
tationally driven fragment approach. Curr. Opin. Drug DiscoVery (33) Taylor, J. D.; Gilbert, P. J.; Williams, M. A.; Pitt, W. R.; Ladbury, DeV. 2005, 8, 355–364.
J. E. Identification of novel fragment compounds targeted against (8) Zartler, E. R.; Shapiro, M. J. Fragonomics: fragment-based drug the pY pocket of v-Src SH2 by computational and NMR screening discovery. Curr. Opin. Chem. Biol. 2005, 9, 366–370.
and thermodynamic evaluation. Proteins 2007, 67, 981–990.
(9) Erlanson, D. A. Fragment-based lead discovery: a chemical update.
Curr. Opin. Biotechnol. 2006, 17, 643–652.
(34) Black, R. D.; Early, T. A.; Roemer, P. B.; Mueller, O. M.; Mogro- (10) Hajduk, P. J. SAR by NMR: putting the pieces together. Mol. Campero, A.; Turner, L. G.; Johnson, G. A. A high-temperature InterVentions 2006, 6, 266–272.
superconducting receiver for nuclear magnetic resonance microscopy.
(11) Johnson, S. L.; Pellecchia, M. Structure- and fragment-based ap- Science 1993, 259, 793–795.
proaches to protease inhibition. Curr. Top. Med. Chem. 2006, 6, 317–
(35) Dalvit, C.; Mongelli, N.; Papeo, G.; Giordano, P.; Veronesi, M.; Moskau, D.; Kummerle, R. Sensitivity improvement in 19F NMR- (12) Leach, A. R.; Hann, M. M.; Burrows, J. N.; Griffen, E. J. Fragment based screening experiments: theoretical considerations and experi- screening: an introduction. Mol. Biosyst. 2006, 2, 430–446.
mental applications. J. Am. Chem. Soc. 2005, 127, 13380–13385.
(13) Bartoli, S.; Fincham, C. I.; Fattori, D. Fragment-based drug design: (36) Gronquist, M.; Meinwald, J.; Eisner, T.; Schroeder, F. C. Exploring combining philosophy with technology. Curr. Opin. Drug DiscoVery uncharted terrain in nature's structure space using capillary NMR DeV. 2007, 10, 422–429.
spectroscopy: 13 steroids from 50 fireflies. J. Am. Chem. Soc. 2005,
(14) Ciulli, A.; Abell, C. Fragment-based approaches to enzyme inhibition.
Curr. Opin. Biotechnol. 2007, 18, 489–496.
(37) Hartshorn, M. J.; Murray, C. W.; Cleasby, A.; Frederickson, M.; (15) Congreve, M.; Murray, C. W.; Carr, R.; Rees, D. C. Fragment-Based Tickle, I. J.; Jhoti, H. Fragment-based lead discovery using X-ray Lead Discovery. Annual Reports in Medicinal Chemistry; Elsevier, crystallography. J. Med. Chem. 2005, 48, 403–413.
Inc.: New York, 2007; pp 431-448.
(38) Seth, P. P.; Miyaji, A.; Jefferson, E. A.; Sannes-Lowery, K. A.; (16) Hajduk, P. J.; Greer, J. A decade of fragment-based drug design: Osgood, S. A.; Propp, S. S.; Ranken, R.; Massire, C.; Sampath, R.; Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 Ecker, D. J.; Swayze, E. E.; Griffey, R. H. SAR by MS: discovery 2007, 50, 5903–5911.
of a new class of RNA-binding small molecules for the hepatitis C (66) Murray, C. W.; Callaghan, O.; Chessari, G.; Cleasby, A.; Congreve, virus: internal ribosome entry site IIA subdomain. J. Med. Chem. M.; Frederickson, M.; Hartshorn, M. J.; McMenamin, R.; Patel, S.; 2005, 48, 7099–7102.
Wallis, N. Application of fragment screening by X-ray crystal- (39) Hofstadler, S. A.; Sannes-Lowery, K. A. Applications of ESI-MS in lography to beta-secretase. J. Med. Chem. 2007, 50, 1116–1123.
drug discovery: interrogation of noncovalent complexes. Nat. ReV. (67) Edwards, P. D.; Albert, J. S.; Sylvester, M.; Aharony, D.; Andisik, D.; Drug DiscoVery 2006, 5, 585–595.
Callaghan, O.; Campbell, J. B.; Carr, R. A.; Chessari, G.; Congreve, (40) Neumann, T.; Junker, H. D.; Keil, O.; Burkert, K.; Ottleben, H.; M.; Frederickson, M.; Folmer, R. H.; Geschwindner, S.; Koether, G.; Gamer, J.; Sekul, R.; Deppe, H.; Feurer, A.; Tomandl, D.; Metz, G.
Kolmodin, K.; Krumrine, J.; Mauger, R. C.; Murray, C. W.; Olsson, Discovery of thrombin inhibitor fragments from chemical microarray L. L.; Patel, S.; Spear, N.; Tian, G. Application of fragment-based lead screening. Lett. Drug Des. DiscoVery 2005, 2, 590–594.
generation to the discovery of novel, cyclic amidine beta-secretase (41) Huber, W.; Mueller, F. Biomolecular interaction analysis in drug inhibitors with nanomolar potency, cellular activity, and high ligand discovery using surface plasmon resonance technology. Curr. Pharm. efficiency. J. Med. Chem. 2007, 50, 5912–5925.
Des. 2006, 12, 3999–4021.
(68) Frederickson, M.; Callaghan, O.; Chessari, G.; Congreve, M.; Cowan, (42) Neumann, T.; Junker, H. D.; Schmidt, K.; Sekul, R. SPR-based S. R.; Matthews, J. E.; McMenamin, R.; Smith, D. M.; Vinkovic, fragment screening: advantages and applications. Curr. Top. Med. M.; Wallis, N. G. Fragment-based discovery of mexiletine derivatives Chem. 2007, 7, 1630–1642.
as orally bioavailable inhibitors of urokinase-type plasminogen (43) Barker, J.; Courtney, S.; Hesterkamp, T.; Ullmann, D.; Whittaker, activator. J. Med. Chem. 2008, 51, 183–186.
M. Fragment screening by biochemical assay. Expert Opin. Drug (69) Card, G. L.; Blasdel, L.; England, B. P.; Zhang, C.; Suzuki, Y.; DiscoVery 2006, 1, 225–236.
Gillette, S.; Fong, D.; Ibrahim, P. N.; Artis, D. R.; Bollag, G.; (44) Hesterkamp, T.; Barker, J.; Davenport, A.; Whittaker, M. Fragment Milburn, M. V.; Kim, S. H.; Schlessinger, J.; Zhang, K. Y. A family based drug discovery using fluorescence correlation: spectroscopy of phosphodiesterase inhibitors discovered by cocrystallography and techniques: challenges and solutions. Curr. Top. Med. Chem. 2007,
scaffold-based drug design. Nat. Biotechnol. 2005, 23, 201–207.
(70) Oltersdorf, T.; Elmore, S. W.; Shoemaker, A. R.; Armstrong, R. C.; (45) Makara, G. M. On sampling of fragment space. J. Med. Chem. 2007,
Augeri, D. J.; Belli, B. A.; Bruncko, M.; Deckwerth, T. L.; Dinges, J.; Hajduk, P. J.; Joseph, M. K.; Kitada, S.; Korsmeyer, S. J.; Kunzer, (46) Teague, S. J.; Davis, A. M.; Leeson, P. D.; Oprea, T. The design of A. R.; Letai, A.; Li, C.; Mitten, M. J.; Nettesheim, D. G.; Ng, S.; leadlike combinatorial libraries. Angew. Chem., Int. Ed. 1999, 38,
Nimmer, P. M.; O'Connor, J. M.; Oleksijew, A.; Petros, A. M.; Reed, J. C.; Shen, W.; Tahir, S. K.; Thompson, C. B.; Tomaselli, K. J.; (47) Hann, M. M.; Oprea, T. I. Pursuing the leadlikeness concept in Wang, B.; Wendt, M. D.; Zhang, H.; Fesik, S. W.; Rosenberg, S. H.
pharmaceutical research. Curr. Opin. Chem. Biol. 2004, 8, 255–263.
An inhibitor of Bcl-2 family proteins induces regression of solid (48) Lepre, C. A. Fragment-based drug discovery using the SHAPES tumours. Nature 2005, 435, 677–681.
method. Expert Opin. Drug DiscoVery 2007, 2, 1555–1566.
(71) Howard, N.; Abell, C.; Blakemore, W.; Chessari, G.; Congreve, M.; (49) Hopkins, A. L.; Groom, C. R.; Alex, A. Ligand efficiency: a useful Howard, S.; Jhoti, H.; Murray, C. W.; Seavers, L. C.; van Montfort, metric for lead selection. Drug DiscoVery Today 2004, 9, 430–431.
R. L. Application of fragment screening and fragment linking to the (50) Abad-Zapatero, C.; Metz, J. T. Ligand efficiency indices as guideposts discovery of novel thrombin inhibitors. J. Med. Chem. 2006, 49,
for drug discovery. Drug DiscoVery Today 2005, 10, 464–469.
(51) Carr, R. A.; Congreve, M.; Murray, C. W.; Rees, D. C. Fragment- (72) Edwards, P. D. Fragment-based drug discovery of carbonic anhydrase based lead discovery: leads by design. Drug DiscoVery Today 2005,
II inhibitors by dynamic combinatorial chemistry. Drug DiscoVery Today 2007, 12, 497–498.
(52) Kuntz, I. D.; Chen, K.; Sharp, K. A.; Kollman, P. A. The maximal (73) Sharpless, K. B.; Manetsch, R. In situ click chemistry: a powerful affinity of ligands. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 9997–
means for lead discovery. Expert Opin. Drug DiscoVery 2007, 1,
(53) Hajduk, P. J. Fragment-based drug design: how big is too big. J. Med. (74) Mocharla, V. P.; Colasson, B.; Lee, L. V.; Roper, S.; Sharpless, K. B.; Chem. 2006, 49, 6972–6976.
Wong, C. H.; Kolb, H. C. In situ click chemistry: enzyme-generated (54) Murray, C. W.; Verdonk, M. L. Entropic Consequences of Linking inhibitors of carbonic anhydrase II. Angew. Chem., Int. Ed. 2004,
Ligands. Fragment-Based Approaches in Drug DiscoVery; Wiley- VCH Verlag GmbH & Co. KGaA: Weinheim, Germany, 2006; pp (75) Cancilla, M. T.; Erlanson, D. A. Tethering: Fragment-Based Drug Discovery by Mass Spectrometry. Mass Spectrometry in Medicinal (55) Leeson, P. D.; Springthorpe, B. The influence of drug-like concepts Chemistry; Wiley-VCH Verlag GmbH & Co.: Weinheim, Germany, on decision-making in medicinal chemistry. Nat. ReV. Drug DiscoVery 2007; pp 305-320.
2007, 6, 881–890.
(76) O'Brien, T.; Fahr, B. T.; Sopko, M. M.; Lam, J. W.; Waal, N. D.; (56) Saxty, G.; Woodhead, S. J.; Berdini, V.; Davies, T. G.; Verdonk, Raimundo, B. C.; Purkey, H. E.; Pham, P.; Romanowski, M. J.
M. L.; Wyatt, P. G.; Boyle, R. G.; Barford, D.; Downham, R.; Garrett, Structural analysis of caspase-1 inhibitors derived from Tethering.
M. D.; Carr, R. A. Identification of inhibitors of protein kinase B Acta Crystallogr., Sect. F: Struct. Biol. Cryst. Commun. 2005, 61,
using fragment-based lead discovery. J. Med. Chem. 2007, 50, 2293–
(77) Fahr, B. T.; O'Brien, T.; Pham, P.; Waal, N. D.; Baskaran, S.; (57) Dill, K. A. Additivity principles in biochemistry. J. Biol. Chem. 1997,
Raimundo, B. C.; Lam, J. W.; Sopko, M. M.; Purkey, H. E.; Romanowski, M. J. Tethering identifies fragment that yields potent (58) Baurin, N.; Aboul-Ela, F.; Barril, X.; Davis, B.; Drysdale, M.; inhibitors of human caspase-1. Bioorg. Med. Chem. Lett. 2006, 16,
Dymock, B.; Finch, H.; Fromont, C.; Richardson, C.; Simmonite, H.; Hubbard, R. E. Design and characterization of libraries of (78) Wood, W. J.; Patterson, A. W.; Tsuruoka, H.; Jain, R. K.; Ellman, molecular fragments for use in NMR screening against protein targets.
J. A. Substrate activity screening: a fragment-based method for the J. Chem. Inf. Comput. Sci. 2004, 44, 2157–2166.
rapid identification of nonpeptidic protease inhibitors. J. Am. Chem. (59) Hubbard, R. E.; Chen, I.; Davis, B. Informatics and modeling Soc. 2005, 127, 15521–15527.
challenges in fragment-based drug discovery. Curr. Opin. Drug (79) Patterson, A. W.; Wood, W. J.; Hornsby, M.; Lesley, S.; Spraggon, DiscoVery DeV. 2007, 10, 289–297.
G.; Ellman, J. A. Identification of selective, nonpeptidic nitrile (60) Schnur, D. M.; Hermsmeier, M. A.; Tebben, A. J. Are target-family- inhibitors of cathepsins using the substrate activity screening method.
privileged substructures truly privileged? J. Med. Chem. 2006, 49,
J. Med. Chem. 2006, 49, 6298–6307.
(80) Soellner, M. B.; Rawls, K. A.; Grundner, C.; Alber, T.; Ellman, J. A.
(61) Congreve, M.; Carr, R.; Murray, C.; Jhoti, H. A "rule of three" for Fragment-based substrate activity screening method for the identifica- fragment-based lead discovery? Drug DiscoVery Today 2003, 8, 876–
tion of potent inhibitors of the Mycobacterium tuberculosis phos- phatase PtpB. J. Am. Chem. Soc. 2007, 129, 9613–9615.
(62) Hann, M. M.; Leach, A. R.; Harper, G. Molecular complexity and (81) Huth, J. R.; Park, C.; Petros, A. M.; Kunzer, A. R.; Wendt, M. D.; its impact on the probability of finding leads for drug discovery.
Wang, X.; Lynch, C. L.; Mack, J. C.; Swift, K. M.; Judge, R. A.; J. Chem. Inf. Comput. Sci. 2001, 41, 856–864.
Chen, J.; Richardson, P. L.; Jin, S.; Tahir, S. K.; Matayoshi, E. D.; (63) Siegal, G.; Ab, E.; Schultz, J. Integration of fragment screening and Dorwin, S. A.; Ladror, U. S.; Severin, J. M.; Walter, K. A.; Bartley, library design. Drug DiscoVery Today 2007, 12, 1032–1039.
D. M.; Fesik, S. W.; Elmore, S. W.; Hajduk, P. J. Discovery and (64) Alex, A. A.; Flocco, M. M. Fragment-based drug discovery: what design of novel HSP90 inhibitors using multiple fragment-based has it achieved so far? Curr. Top. Med. Chem. 2007, 7, 1544–1567.
design strategies. Chem. Biol. Drug Des. 2007, 70, 1–12.
(65) Geschwindner, S.; Olsson, L. L.; Albert, J. S.; Deinum, J.; Edwards, (82) Vieth, M.; Siegel, M. G.; Higgs, R. E.; Watson, I. A.; Robertson, P. D.; Beer, T.; Folmer, R. H. Discovery of a novel warhead against D. H.; Savin, K. A.; Durst, G. L.; Hipskind, P. A. Characteristic beta-secretase through fragment-based lead generation. J. Med. Chem. physical properties and structural fragments of marketed oral drugs.
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 3679
J. Med. Chem. 2004, 47, 224–232.
(104) Baxter, C. A.; Murray, C. W.; Waszkowycz, B.; Li, J.; Sykes, R. A.; (83) Lepre, C. A.; Moore, J. M.; Peng, J. W. Theory and applications of Bone, R. G.; Perkins, T. D.; Wylie, W. New approach to molecular NMR-based screening in pharmaceutical research. Chem. ReV. 2004,
docking and its application to virtual screening of chemical databases.
J. Chem. Inf. Comput. Sci. 2000, 40, 254–262.
(84) Lewell, X. Q.; Judd, D. B.; Watson, S. P.; Hann, M. M.
(105) Jones, G.; Willett, P.; Glen, R. C.; Leach, A. R.; Taylor, R.
RECAPsretrosynthetic combinatorial analysis procedure: a powerful Development and validation of a genetic algorithm for flexible new technique for identifying privileged molecular fragments with docking. J. Mol. Biol. 1997, 267, 727–748.
useful applications in combinatorial chemistry. J. Chem. Inf. Comput. (106) Kramer, B.; Rarey, M.; Lengauer, T. Evaluation of the FLEXX Sci. 1998, 38, 511–522.
incremental construction algorithm for protein-ligand docking.
(85) Kolb, P.; Caflisch, A. Automatic and efficient decomposition of two- Proteins 1999, 37, 228–241.
dimensional structures of small molecules for fragment-based high- (107) Pang, Y. P.; Perola, E.; Xu, K.; Prendergast, F. G. EUDOC: a throughput docking. J. Med. Chem. 2006, 49, 7384–7392.
computer program for identification of drug interaction sites in (86) Warren, G. L.; Andrews, C. W.; Capelli, A. M.; Clarke, B.; LaLonde, macromolecules and drug leads from chemical databases. J. Comput. J.; Lambert, M. H.; Lindvall, M.; Nevins, N.; Semus, S. F.; Senger, Chem. 2001, 22, 1750–1771.
S.; Tedesco, G.; Wall, I. D.; Woolven, J. M.; Peishoff, C. E.; Head, (108) Nissink, J. W.; Murray, C.; Hartshorn, M.; Verdonk, M. L.; Cole, M. S. A critical assessment of docking programs and scoring J. C.; Taylor, R. A new test set for validating predictions of protein- functions. J. Med. Chem. 2006, 49, 5912–5931.
ligand interaction. Proteins 2002, 49, 457–471.
(87) Vangrevelinghe, E.; Rudisser, S. Computational approaches for (109) Hartshorn, M. J.; Verdonk, M. L.; Chessari, G.; Brewerton, S. C.; fragment optimization. Curr. Comput.-Aided Drug Des. 2007, 3, 69–
Mooij, W. T.; Mortenson, P. N.; Murray, C. W. Diverse, high-quality test set for the validation of protein-ligand docking performance.
(88) Bohm, H. J.; Banner, D. W.; Weber, L. Combinatorial docking and J. Med. Chem. 2007, 50, 726–741.
combinatorial chemistry: design of potent non-peptide thrombin (110) Hartshorn, M. J. AstexViewer: a visualisation aid for structure-based inhibitors. J. Comput.-Aided Mol. Des. 1999, 13, 51–56.
drug design. J. Comput.-Aided Mol. Des. 2002, 16, 871–881.
(89) Liebeschuetz, J. W.; Jones, S. D.; Morgan, P. J.; Murray, C. W.; (111) Selinsky, B. S.; Gupta, K.; Sharkey, C. T.; Loll, P. J. Structural Rimmer, A. D.; Roscoe, J. M.; Waszkowycz, B.; Welsh, P. M.; analysis of NSAID binding by prostaglandin H2 synthase: time- Wylie, W. A.; Young, S. C.; Martin, H.; Mahler, J.; Brady, L.; dependent and time-independent inhibitors elicit identical enzyme Wilkinson, K. PRO_SELECT: combining structure-based drug design conformations. Biochemistry 2001, 40, 5172–5180.
and array-based chemistry for rapid lead discovery. 2. The develop- (112) Hajduk, P. J.; Boyd, S.; Nettesheim, D.; Nienaber, V.; Severin, J.; ment of a series of highly potent and selective factor Xa inhibitors.
Smith, R.; Davidson, D.; Rockway, T.; Fesik, S. W. Identification J. Med. Chem. 2002, 45, 1221–1232.
of novel inhibitors of urokinase via NMR-based screening. J. Med. (90) Kick, E. K.; Roe, D. C.; Skillman, A. G.; Liu, G.; Ewing, T. J.; Sun, Chem. 2000, 43, 3862–3866.
Y.; Kuntz, I. D.; Ellman, J. A. Structure-based design and combi- (113) Huth, J. R.; Sun, C.; Sauer, D. R.; Hajduk, P. J. Utilization of NMR- natorial chemistry yield low nanomolar inhibitors of cathepsin D.
derived fragment leads in drug design. Methods Enzymol. 2005, 394,
Chem. Biol. 1997, 4, 297–307.
549–571. (Nuclear Magnetic Resonance of Biological Macromol-ecules, Part C) (91) Gradler, U.; Gerber, H. D.; Goodenough-Lashua, D. M.; Garcia, G. A.; Ficner, R.; Reuter, K.; Stubbs, M. T.; Klebe, G. A new target (114) Warnmark, A.; Treuter, E.; Gustafsson, J. A.; Hubbard, R. E.; for shigellosis: rational design and crystallographic studies of Brzozowski, A. M.; Pike, A. C. Interaction of transcriptional inhibitors of tRNA-guanine transglycosylase. J. Mol. Biol. 2001, 306,
intermediary factor 2 nuclear receptor box peptides with the coac- tivator binding site of estrogen receptor alpha. J. Biol. Chem. 2002,
277, 21862–21868.
(92) Honma, T.; Hayashi, K.; Aoyama, T.; Hashimoto, N.; Machida, T.; (115) Brzozowski, A. M.; Pike, A. C.; Dauter, Z.; Hubbard, R. E.; Bonn, Fukasawa, K.; Iwama, T.; Ikeura, C.; Ikuta, M.; Suzuki-Takahashi, T.; Engstrom, O.; Ohman, L.; Greene, G. L.; Gustafsson, J. A.; I.; Iwasawa, Y.; Hayama, T.; Nishimura, S.; Morishima, H. Structure- Carlquist, M. Molecular basis of agonism and antagonism in the based generation of a new class of potent Cdk4 inhibitors: new de oestrogen receptor. Nature 1997, 389, 753–758.
novo design strategy and library design. J. Med. Chem. 2001, 44,
(116) Rasmussen, H. B.; Branner, S.; Wiberg, F. C.; Wagtmann, N. Crystal structure of human dipeptidyl peptidase IV/CD26 in complex with a (93) Marcou, G.; Rognan, D. Optimizing fragment and scaffold docking substrate analog. Nat. Struct. Biol. 2003, 10, 19–25.
by use of molecular interaction fingerprints. J. Chem. Inf. Model. (117) Edmondson, S. D.; Mastracchio, A.; Mathvink, R. J.; He, J.; Harper, 2007, 47, 195–207.
B.; Park, Y. J.; Beconi, M.; Di Salvo, J.; Eiermann, G. J.; He, H.; (94) Verdonk, M. L.; Berdini, V.; Hartshorn, M. J.; Mooij, W. T.; Murray, Leiting, B.; Leone, J. F.; Levorse, D. A.; Lyons, K.; Patel, R. A.; C. W.; Taylor, R. D.; Watson, P. Virtual screening using protein- Patel, S. B.; Petrov, A.; Scapin, G.; Shang, J.; Roy, R. S.; Smith, A.; ligand docking: avoiding artificial enrichment. J. Chem. Inf. Comput. Wu, J. K.; Xu, S.; Zhu, B.; Thornberry, N. A.; Weber, A. E. (2S,3S)- Sci. 2004, 44, 793–806.
(95) Clark, M.; Guarnieri, F.; Shkurko, I.; Wiseman, J. Grand Canonical Monte Carlo Simulation of Ligand-Protein Binding. J. Chem. Inf. amino amide dipeptidyl peptidase IV inhibitor for the treatment of Model. 2007, 46, 231–242.
type 2 diabetes. J. Med. Chem. 2006, 49, 3614–3627.
(96) Rummey, C.; Nordhoff, S.; Thiemann, M.; Metz, G. In silico (118) Alderton, W. K.; Cooper, C. E.; Knowles, R. G. Nitric oxide fragment-based discovery of DPP-IV S1 pocket binders. Bioorg. Med. synthases: structure, function and inhibition. Biochem. J. 2001, 357,
Chem. Lett. 2006, 16, 1405–1409.
(97) Nordhoff, S.; Cerezo-Galvez, S.; Feurer, A.; Hill, O.; Matassa, V. G.; (119) Fedorov, R.; Hartmann, E.; Ghosh, D. K.; Schlichting, I. Structural Metz, G.; Rummey, C.; Thiemann, M.; Edwards, P. J. The reversed basis for the specificity of the nitric-oxide synthase inhibitors W1400 binding of beta-phenethylamine inhibitors of DPP-IV: X-ray struc- and Nomega-propyl-L-Arg for the inducible and neuronal isoforms.
tures and properties of novel fragment and elaborated inhibitors.
J. Biol. Chem. 2003, 278, 45818–45825.
Bioorg. Med. Chem. Lett. 2006, 16, 1744–1748.
(120) Tisi, D.; Chessari, G.; Woodhead, A. J.; Jhoti, H. Structural Biology (98) Warner, S. L.; Bashyam, S.; Vankayalapati, H.; Bearss, D. J.; Han, and Anticancer Drug Design. Cancer Drug Design and DiscoVery; H.; Mahadevan, D.; Von Hoff, D. D.; Hurley, L. H. Identification of Elsevier: New York, 2008; pp 91-106.
a lead small-molecule inhibitor of the Aurora kinases using a (121) Grunewald, G. L.; Romero, F. A.; Criscione, K. R. 3-Hydroxymethyl- structure-assisted, fragment-based approach. Mol. Cancer Ther. 2006,
tors of phenylethanolamine N-methyltransferase that display remark- (99) Babaoglu, K.; Shoichet, B. K. Deconstructing fragment-based inhibi- able potency and selectivity. J. Med. Chem. 2005, 48, 134–140.
tor discovery. Nat. Chem. Biol. 2006, 2, 720–723.
(122) Kallander, L. S.; Lu, Q.; Chen, W.; Tomaszek, T.; Yang, G.; Tew, (100) Velentza, A. V.; Wainwright, M. S.; Zasadzki, M.; Mirzoeva, S.; D.; Meek, T. D.; Hofmann, G. A.; Schulz-Pritchard, C. K.; Smith, Schumacher, A. M.; Haiech, J.; Focia, P. J.; Egli, M.; Watterson, W. W.; Janson, C. A.; Ryan, M. D.; Zhang, G. F.; Johanson, K. O.; D. M. An aminopyridazine-based inhibitor of a pro-apoptotic protein Kirkpatrick, R. B.; Ho, T. F.; Fisher, P. W.; Mattern, M. R.; Johnson, kinase attenuates hypoxia-ischemia induced acute brain injury.
R. K.; Hansbury, M. J.; Winkler, J. D.; Ward, K. W.; Veber, D. F.; Bioorg. Med. Chem. Lett. 2003, 13, 3465–3470.
Thompson, S. K. 4-Aryl-1,2,3-triazole: a novel template for a (101) Keseru, G. M.; Makara, G. M. Hit discovery and hit-to-lead reversible methionine aminopeptidase 2 inhibitor, optimized to inhibit approaches. Drug DiscoVery Today 2006, 11, 741–748.
angiogenesis in vivo. J. Med. Chem. 2005, 48, 5644–5647.
(102) Wells, J. A.; McClendon, C. L. Reaching for high-hanging fruit in (123) Marino, J. P., Jr.; Fisher, P. W.; Hofmann, G. A.; Kirkpatrick, R. B.; drug discovery at protein-protein interfaces. Nature 2007, 450, 1001–
Janson, C. A.; Johnson, R. K.; Ma, C.; Mattern, M.; Meek, T. D.; Ryan, M. D.; Schulz, C.; Smith, W. W.; Tew, D. G.; Tomazek, T. A., (103) Hajduk, P. J. Puzzling through fragment-based drug design. Nat. Jr.; Veber, D. F.; Xiong, W. C.; Yamamoto, Y.; Yamashita, K.; Yang, Chem. Biol. 2006, 2, 658–659.
G.; Thompson, S. K. Highly potent inhibitors of methionine ami- Journal of Medicinal Chemistry, 2008, Vol. 51, No. 13 nopeptidase-2 based on a 1,2,4-triazole pharmacophore. J. Med. Chem. 2007, 50, 3777–3785.
(127) Hajduk, P. J.; Huth, J. R.; Fesik, S. W. Druggability indices for protein (124) Brenk, R.; Meyer, E. A.; Reuter, K.; Stubbs, M. T.; Garcia, G. A.; targets derived from NMR-based screening data. J. Med. Chem. 2005,
Diederich, F.; Klebe, G. Crystallographic study of inhibitors of tRNA- guanine transglycosylase suggests a new structure-based pharma- (128) Congreve, M.; Murray, C. W.; Blundell, T. L. Structural biology cophore for virtual screening. J. Mol. Biol. 2004, 338, 55–75.
and drug discovery. Drug DiscoVery Today 2005, 10, 895–907.
(125) Stengl, B.; Meyer, E. A.; Heine, A.; Brenk, R.; Diederich, F.; Klebe, (129) Albert, J. S.; Blomberg, N.; Breeze, A. L.; Brown, A. J.; Burrows, G. Crystal structures of tRNA-guanine transglycosylase (TGT) in J. N.; Edwards, P. D.; Folmer, R. H.; Geschwindner, S.; Griffen, complex with novel and potent inhibitors unravel pronounced E. J.; Kenny, P. W.; Nowak, T.; Olsson, L. L.; Sanganee, H.; Shapiro, induced-fit adaptations and suggest dimer formation upon substrate A. B. An integrated approach to fragment-based lead generation: binding. J. Mol. Biol. 2007, 370, 492–511.
philosophy, strategy and case studies from AstraZeneca's drug (126) Congreve, M.; Aharony, D.; Albert, J.; Callaghan, O.; Campbell, J.; discovery programmes. Curr. Top. Med. Chem. 2007, 7, 1600–1629.
Carr, R. A.; Chessari, G.; Cowan, S.; Edwards, P. D.; Frederickson, (130) Sprang, S. R. Structural biology: a receptor unlocked. Nature 2007,
M.; McMenamin, R.; Murray, C. W.; Patel, S.; Wallis, N. Application of fragment screening by X-ray crystallography to the discovery of
aminopyridines as inhibitors of beta-secretase. J. Med. Chem. 2007,

Source: ftp://ftp.essex.ac.uk/pub/oyster/BS317/lectures/Congreve_JMC_FBDD_rev_08.pdf

Capitolato prestaz.special d'app grangia

Comune di Orio Litta Progetto di Restauro "Ostel o Grangia dei Benedettini" Febbraio 2012 Comune di Orio Litta – Intervento di Restauro/ Risanamento Conservativo, Riqualificazione Funzionale e Adeguamento Impianti dell'Ostello dei Pellegrini della Via Francigena " Grangia dei Benedettini" DISCIPLINARE DESCRITTIVO E PRESTAZIONALE DEGLI ELEMENTI TECNICI


Udredningsopgave for Fødevarestyrelsen Kostens betydning for læring og adfærd hos børn En gennemgang af den videnskabelige litteratur Institut for Human Ernæring Det Biovidenskabelig Fakultet Københavns Universitet Kostens betydning for læring og adfærd hos børn 1. Forord _ 3 2.

Copyright © 2008-2016 No Medical Care