The importance of specific reactivity for covalent drugs

As we noted in our thousandth post, covalent drugs are becoming increasingly popular, particularly for tackling tough targets. But finding and optimizing covalent ligands entails unique challenges, as discussed in a new paper by Bharath Srinivasan at Cancer Research UK. (Derek Lowe also recently blogged about this.)

 

Interactions between noncovalent drugs and their targets are characterized by dissociation or inhibition constants K_D or K_{I ,} where lower numbers mean stronger binding. In contrast, irreversible covalent drugs are characterized by a ratio we discussed last year, k_inact/K_I, where the rate constant k_inact represents the covalent modification step. (Side note: although the term k_inact is commonly used, covalent modulators can also be activators; my company Frontier Medicines recently announced a covalent activator of p53^Y220C. Perhaps k_cov would be more general?)

 

To explain k_inact/K_I, Srinivasan draws a useful analogy to enzymes, which are mechanistically described by the specificity constant k_cat/K_m in Michaelis-Menten kinetics. In both cases, higher numbers mean more rapid modification or greater catalytic efficiency. A study of several thousand enzymes found the median k_cat/K_m to be around 100,000 M^-1s^-1, with 60% between 1,000 and 1,000,000 M^-1s^-1. Enzymes operate by stabilizing the transition state of the reaction, which means that the affinities for the substrates do not necessarily have to be high, particularly if the structures of the substrates differ from the transition states.

 

Just as catalytic efficiency for enzymes can be increased either by increasing k_cat or lowering K_m, the inactivation efficiency of covalent drugs can be optimized either by increasing k_inact or by decreasing K_I. Historically, drug hunters have focused on the latter; we previously described the discovery of TAK-020 in which the affinity of a fragment for the kinase BTK was first optimized and then a covalent warhead was appended.

 

However, focusing on k_inact can also be productive, and Srinivasan argues this is particularly true for challenging targets with shallow pockets where noncovalent affinity is difficult to obtain. As a case in point he discusses covalent KRAS^G12C inhibitors such as sotorasib, which I wrote about here. Just as residues within enzyme active sites stabilize the transition state of a reaction, a lysine residue in KRAS forms a hydrogen bond to the carbonyl of the acrylamide electrophile, thereby increasing its reactivity for the protein.

 

Srinivasan emphasizes that k_inact is specific for each particular protein-ligand pair as well as distinct from intrinsic or chemical reactivity. This is a critical point. Newcomers to the field often worry that a high k_inact value means a molecule is generically reactive and thus likely to react with many proteins, but this is not necessarily true. For example, sotorasib’s favorable k_inact/K_I is driven by a high k_inact for KRAS^G12C but it is still quite specific. Indeed, Srinivasan points out that even a chemically reactive molecule may not react with a protein if the geometry isn’t right.

 

A nice way of assessing specific reactivity (which unfortunately is not cited) is the reactivity enhancement factor, or REF, as defined by Alan Armstrong, David Mann, and colleagues at Imperial College London in an (open-access) 2020 ChemBioChem paper. Akin to the k_cat/k_uncat ratio used to assess rate enhancement for enzymes, REF is defined as the rate of reaction for a specific protein divided by the rate of reaction for glutathione, an abundant cellular thiol. The higher the REF score, the higher the specific reactivity for the protein of interest.

 

Srinivasan also considers tradeoffs between k_inact and K_I as k_inact/K_I approaches the rate of diffusion, suggesting that above 1,000,000 M^-1s^-1 or so any further improvement in affinity will come at the cost of specific reactivity. While this is theoretically interesting, from a practical perspective you can have a perfectly fine drug with a k_inact/K_I of just 10,000 M^-1s^-1.

 

Covalent drugs will only become more important as we pursue increasingly hard targets that have resisted previous efforts. For these targets in particular, focusing on specific reactivity will be rewarding.

An update on the COVID Moonshot

On March 18, 2020, a group called the COVID Moonshot released crystal structures of 71 fragments bound to the SARS-CoV2 Mpro protein. The same day, they launched an online crowdsourcing initiative seeking ideas for how to advance these fragments, none of which had activity in an enzymatic assay. The results of this experiment in open science have just been published in Science, appropriately open-access.

 

Within the first week, the group received more than 2000 submissions. Ultimately more than 20,000 molecules were submitted, and all of these were evaluated in “alchemical free-energy calculations.” These are computationally intensive, requiring ~80 GPU hours per compound, so the consortium used the volunteer-based distributed computing network Folding@home. Compounds were evaluated not just for potency but also synthetic accessibility, and those that passed were synthesized at Enamine and tested in various functional assays.

 

In addition to accepting submissions for how to advance fragments, a core group of researchers proposed their own ideas. Interestingly, at least in the early stages of the project, this elite group did no better at coming up with more potent or synthetically accessible molecules, despite being intimately involved with the project. This finding validates the open-sourcing of ideas from the larger scientific community.

 

Ultimately more than 2400 compounds were synthesized, and more than 500 crystal structures were determined. All experimental results were posted online to help guide the synthesis of additional compounds. Speed was consistently prioritized, not just with high-throughput crystallography but also high-throughput chemistry and "direct-to-biology" screening of crude reaction mixtures.

 

The paper highlights one lead series, which originated from a community submission (TRY-UNI-714a760b-6, itself fragment-sized) inspired by merging overlapping fragments. This mid micromolar inhibitor was ultimately optimized to MAT-POS-e194df51-1, with mid-nanomolar activity in both biochemical and cell assays. (Despite a chloroacetamide in one of the original fragments and a nitrile in the final molecule, which is the warhead found in the approved covalent Mpro inhibitor nirmatrelvir, MAT-POS-e194df51-1 is non-covalent.) 

 

The molecule is potent against known SARS-CoV-2 variants, including recent ones such as Omicron. A crystal structure of the final molecule also overlays remarkably well onto the initial fragments.

 

The paper notes that there is still considerable work to do, particularly optimizing the pharmacokinetics to lower clearance and improve bioavailability. These efforts can take vast sums of time and money, and the lead series has been adopted by the Drugs for Neglected Diseases initiative for further development. Although a handful of drugs are already approved against SARS-CoV-2, there is room for improvement: Derek Lowe posted a vivid personal account of his experience on nirmatrelvir here.

 

When we wrote about the COVID Moonshot in March of 2020, we correctly predicted that vaccines would be approved before drugs from this effort emerged. Fortunately, our warning that “there will be a SARS-CoV-3” has not proven correct – yet. But open science endeavors such as the COVID Moonshot will help us prepare for this eventuality. We may not have made it to the moon yet, but perhaps we’ve learned how to leave Earth’s orbit.

Crude reaction screening by crystallography

Getting from a fragment to something more useful can be time consuming, and frustrating. Often it requires making lots of inactive analogs. It is easy to set up multiple chemical reactions in parallel, but purifying products is tedious. Thus, there has been a trend towards screening crude reaction mixtures. Off-rate screening (ORS) using SPR was published back in 2014, and other techniques have been used as well. Now, two recent papers describe using crystallography as the assay.

 

In a J. Med. Chem. paper, Bradley Doak, Martin Scanlon, and collaborators at Monash University, La Trobe University, University of Wollongong, and Queensland University of Technology describe Rapid Elaboration of Fragments into Leads using X-ray crystallography (REFiL_x; see also here). The researchers note that ORS is best for identifying reasonably potent compounds, with low (<10) micromolar potency or better. For some targets, this is a bar too high. Specifically, they were interested in E. coli disulfide bond forming protein A (EcDsbA), an anti-virulence target for which they had found – after considerable effort – a 490 µM fragment.

 

The fragment contained a carboxylic acid, which could be conveniently used for amide bond formation, and the researchers decided to make a bespoke library around it. A collection of 93 small (5-12 heavy atoms) amines was assembled and each member was reacted with 2 micromoles (about 0.5 mg) of the fragment in a 96 well plate. The reactions were then evaporated and dissolved in DMSO to 100 mM concentration (assuming the reactions went to completion). HPLC-MS analyses revealed likely product for 83 reactions with yields up to 90%, though the average was closer to 25%. These crude reaction mixtures were then soaked into previously grown crystals of EcDsbA and analyzed crystallographically using both manual and automatic processing (including PanDDA).

 

The result was four hits, all of which were resynthesized and characterized in detail. Crystallography of the pure compounds confirmed the binding modes found in the crude reaction mixtures. Gratifyingly, one of the compounds bound 8-fold more tightly than the initial fragment, as assessed by two dimensional NMR.

 

The four hits all had purities >50% in the crude reaction mixtures, so the researchers resynthesized a few other non-hits that had lower yields. One of these was about 2-fold better than the original fragment, suggesting a false negative in the first screen. Experiments in which the most active hit was spiked into faux reaction mixtures at increasingly lower concentrations revealed – as expected – that detection was more difficult at lower concentrations.

 

A similar approach is described in Communications Chemistry by Rod Hubbard and collaborators at Vernalis, University of York, Diamond Light Source, University of Oxford, and University of Johannesburg. However, whereas the first paper focuses on trying to find improved hits against a difficult target, this paper focuses more on methodology. The researchers used two of the same protein targets for which they had previously performed off-rate screening, HSP90 and PDHK2. Conveniently, both these ATPases share some inhibitors, so the same set of 83 crude reaction mixtures could be used for both enzymes. Also in contrast to the EcDsbA example, the starting fragments (found in prior screens) were considerably more potent, as were the final molecules.

 

The researchers note that “as with any high-throughput experimental approach, there are false positives and false negatives.” Because off-rates were measured for all the crude reaction mixtures, the researchers could assess false positives and false negatives in the crystallographic screens. In false positives, products were seen in the crystal structures even though their affinities were no better than the reactant. In false negatives, starting material was seen in the crystal structures even though the products had better affinities. Together, these accounted for about half the results. Digging into the details, poor reaction yields and low solubility for the products accounted for many of the false negatives. False positives are harder to explain, though the researchers note that the buffer used for crystal soaking is different from that used for SPR.

 

There were also some notable successes: for one racemic compound, crystals of PDHK2 “correctly” selected the more potent enantiomer (K_d = 0.14 µM) over the less potent one (K_d = 17 µM). As in the first paper, there is a wealth of experimental details, including improvements to PanDDA protocols. The researchers performed the soaks in triplicate, and although this did lead to an increased number of hits, they note that singleton screening would probably be sufficient for most applications.

 

At the CHI FBDD meeting last month Frank von Delft, who is an author of the second paper, noted that he is increasingly using crude reaction screening to progress fragments, including against SARS-CoV-2. I look forward to seeing this approach, and the leads that come from it, advance.

Fragment merging on the WBM site of scaffold protein WDR5

Two years ago we highlighted work out of Stephen Fesik’s lab at Vanderbilt University describing potent binders of WDR5, a molecular scaffold that interacts with dozens of other proteins. Those molecules bind at the so-called WIN site, disrupting interactions with proteins such as MLL1. Other proteins, such as the famous anticancer target MYC, bind at a completely different location – the WBM site. This is the focus of a new paper from the same group in J. Med. Chem.

The researchers had previously completed a traditional high-throughput screen and identified molecules such as compound 1. These were further optimized, but, as one might expect looking at the chemical structure, the best molecules had “challenging physicochemical profiles.” The researchers turned to fragments for help.

A two-dimensional (¹H-¹⁵N HMQC) NMR screen of ~14,000 fragments yielded 43 hits, all of them quite weak, with dissociation constants in the millimolar range. The tetrapeptide portion of MYC that binds to the WBM site, Ile-Asp-Val-Val, contains a carboxylic acid flanked by lipophilic residues, and as one would expect many hits were hydrophobic acids. Crystal structures were determined for five, and these suggested a fragment merging opportunity.

The carboxylic acid moiety of fragment F2 makes similar interactions with an asparagine residue in WBM as the sulfonamide moiety of compound 1. The resulting merged compound 2a showed improved potency. More than a dozen replacements for the cyclohexyl ring were attempted but none improved potency significantly. Similarly, moving the cycloalkyl group around the 5-membered heterocycle was not productive. However, introducing a methyl sulfone moiety to engage a lysine residue led to a ten-fold boost in potency for compound 12. The molecule disrupted WDR5-MYC complex formation in cell lysates and also reduced MYC binding to target genes in cells.

This is another nice example of using fragment merging to fix problems across early lead series. Of course, compound 12 still has a long way to go; as the researchers note, the phenol is a likely site of glucuronidation. Still, this and the 2018 paper demonstrate the power of fragments to target two separate protein-protein interfaces on the same protein.

Machines, fixing human disease

Last year we highlighted the secretive juggernaut DREADCO's move into drug discovery. Today they announced the launch of their new division SkyFragNet (not to be confused with the European graduate training program FragNet). Its audacious mission: “to eradicate human disease."

SkyFragNet will automate every aspect of drug discovery. The approach starts with a powerful docking method, in which all 166 billion members of GDB-17 will be docked against a target of interest. Synthetic schemes for the virtual hits will be computationally generated, and the compounds will be synthesized using automated flow synthesis and mass-directed purification.

Fragment hits that confirm in a panel of biophysical techniques will then undergo computational-based growing; SkyFragNet incorporates the latest AI algorithms to maximize the likelihood of success. As with the fragments, designed molecules will be synthesized and tested, first in biochemical and then in cell-based assays.

Although the folks at Mordor State College are trying to make animal testing obsolete, SkyFragNet will still rely on pharmaokinetic and pharmacodynamic studies. However, they have built a fully mechanized vivarium run entirely by robots - think of The Matrix but with mice in place of humans.

Finally, compounds that make it through this gauntlet will be scaled up under GMP conditions (automated, of course) for clinical trials. It remains to be seen how many compounds SkyFragNet will take into the clinic, or whether the success rates will be higher than those of their human counterparts.

Of course, with all this power comes enormous responsibility. If things go wrong, hopefully DREADCO will have the wisdom to Terminate the program. Eradicating human disease could be done in two very different ways.

Review of 2018 reviews

As 2018 recedes into history, we are using this last post of the year to do what we have done since 2012 – review notable events along with reviews we didn’t previously cover.

This was a busy year for meetings, starting in January with a FragNet event in Barcelona, then moving to San Diego in April for the annual CHI FBDD meeting. Boston saw an embarrassment of riches, from the first US-based NovAliX meeting, to a symposium on FBDD at the Fall ACS meeting, followed closely by a number of relevant talks at CHI’s Discovery on Target. Finally, the tenth anniversary of the renowned FBLD meeting returned to San Diego. Look for a schedule of 2019 events later this month.

If meetings were abundant, the same can be said for reviews.

Lead optimization

Writing in J. Med. Chem., Dean Brown and Jonas Boström (AstraZeneca) asked “where do recent small molecule clinical development candidates come from?” For three quarters of the 66 molecules published in J. Med. Chem. in 2016 and 2017 the answer is from known compounds or HTS, though fragments accounted for four examples. Although average molecular weight increased during lead optimization, lipophilicity did not, suggesting the importance of this parameter.

The importance of keeping lipophilicity in check is also emphasized by Robert Young (GlaxoSmithKline) and Paul Leeson (Paul Leeson Consulting) in a massive J. Med. Chem. treatise on lead optimization. Buttressed with dozens of examples, including several from FBLD, they show that the final molecule is usually among the most efficient (in terms of LE and LLE) in a given series, even when metrics were not explicitly used by the project team. Perhaps with pedants like Dr. Saysno in mind, they also emphasize the complexity of drug discovery, and note that “seeking optimum efficiencies and physicochemical properties are guiding principles and not rules.”

Lipophilic ligand efficiency (LLE) is also the focus of a paper in Bioorg. Med. Chem. by James Scott (AstraZeneca) and Michael Waring (Newcastle University). This is based largely on personal experiences and provides lots of helpful tips. Importantly, the researchers note that calculated lipophilicity values can differ dramatically from measured values, and go so far as to say that “this variation is sufficient to render LLEs derived from calculated values meaningless.”

Turning wholly to fragments, Chris Johnson and collaborators (including yours truly) from Astex, Carmot, Vrije Universiteit Amsterdam, and Novartis have published an analysis in J. Med. Chem. of fragment-to-lead success stories from last year. This review, the third in a series, also summarizes all 85 examples published between 2015 and 2017, confirming and expanding some of the trends we mentioned last year.

Targets

Two reviews focus on specific target classes. Bas Lamoree and Rod Hubbard (University of York) cover antibiotics in SLAS Discovery. After a nice, concise review of fragment-finding methods, the researchers discuss a number of case studies, many of which will be familiar to regular readers of this blog, including an early example of whole-cell screening.

David Bailey and collaborators from IOTA and University of Cambridge discuss cyclic nucleotide phosphodiesterases (PDEs) in J. Med. Chem. The researchers provide a good overview of the field, including mining the open database ChEMBL for fragment-sized inhibitors. As they point out, the first inhibitors discovered for these cell-signaling enzymes were fragment-sized, so it is no surprise that FBLD has been fruitful – see here for an example from earlier this year. Interestingly though, although at least six fragment-sized PDE inhibitor drugs have been approved, none of these were actually discovered using FBLD.

PDEs are an example of “ligandable” targets, for which small molecule modulators are readily discovered. In Drug Discovery Today, Sinisa Vukovic and David Huggins (University of Cambridge) discuss ligandability “in terms of the balance between effort and reward.” They use a published database of protein-ligand affinities to develop a metric, LIG_exp, for experimental ligandability, and also describe their computational metric, Solvaware, which is based on identifying clusters of water molecules binding weakly to a protein. Comparisons with experimental data and with other predictive metrics, such as FTMap, reveal that while the computational methods are useful, there is still room for improvement.

We have previously written about how target-guided synthesis methods such as dynamic combinatorial chemistry have – despite decades of research – yielded few truly novel, drug-like ligands. Is this because the targets chosen were simply not ligandable? In J. Med. Chem., Anna Hirsch and collaborators at the University of Groningen, the Helmholtz Institute for Pharmaceutical Research, and Saarland University review some (though by no means all) published examples and examine their computationally determined ligandability scores. There seems to be no difference between these targets and a set of traditional drug targets.

Finding fragments

Crystallography continues to be a key tool for FBLD: as we noted in the review of the 2017 literature, 21 of the 30 examples made use of a crystal structure of either the starting fragment or an analog, and only 3 projects didn’t use crystallography at all. That said, FBLD is possible without crystallography, as illustrated through multiple examples in a Cell Chem. Biol. review by Wolfgang Jahnke (Novartis), Ben Davis (Vernalis), and me (Carmot).

In the absence of a crystal structure, NMR is best suited for providing structural information, and this is the subject of a review in Molecules by Barak Akabayov and colleagues at Ben-Gurion University of the Negev. The researchers provide a nice summary of NMR screening methods and success stories within a broader history of FBLD. They also include an extensive list of fragment library providers as well as a discussion of virtual screening.

Speaking of virtual screening, three reviews cover this topic. In Methods Mol. Biol., Durai Sundar and colleagues at Indian Institute of Technology Delhi touch on a number of computational approaches for de novo ligand design, though the lack of structures sometimes makes it challenging to read. A broader, more visually appealing review is published in AAPS Journal by Yuemin Bian and Xiang-Qun Xie at University of Pittsburgh. In addition to an overview and case studies, the researchers also provide a nice table summarizing 15 different computational programs. One of these, SEED, is a main focus of a review in Eur. J. Med. Chem. by Jean-Rémy Marchand and Amedeo Caflisch (University of Zürich). The researchers describe how this docking program can be combined with X-ray crystallography (SEED2XR) to rapidly identify fragments; we highlighted an example with a bromodomain. Their ALTA protocol uses SEED to generate larger, more potent molecules, as we described for the kinase EphB4. The researchers note that together these protocols have led to about 200 protein-ligand crystal structures deposited in the PDB over the past five years.

Rounding out methods, Sten Ohlson and Minh-Dao Duong-Thi (Nanyang Technological University) provide a detailed how-to guide in Methods for performing weak affinity chromatography, and how this can be combined with mass spectrometry (WAC-MS), as we noted last year.

Chemistry

One drawback of some computational approaches for fragment optimization is that they do not consider synthetic accessibility. In Mol. Inform., Philippe Roche, Xavier Morelli, and collaborators at Aix-Marseille University and Institut Paoli-Calmettes focus on hit to lead approaches that do, and provide a handy table summarizing nearly a dozen computational methods. We highlighted one from the authors, DOTS, earlier this year.

DOTS is an example of using DOS, or diversity-oriented synthesis. In Front. Chem., David Spring and colleagues at University of Cambridge review recent applications of DOS for generating new fragments, some of which we recently highlighted. Only a couple examples of successfully screening these new fragments are described, but the authors note that this is likely to increase as virtual library screening continues to advance.

Perhaps the most productive fragment of all time is 7-azaindole, the origin of three fragment-derived clinical compounds. (The moiety appears in both approved FBLD-derived drugs, vemurafenib and venetoclax.) Takayuki Irie and Masaaki Sawa of Carna Biosciences devote their attention to this little bicycle in Chem. Pharm. Bull. The researchers count six clinical kinase inhibitors that contain 7-azaindole (not all from FBLD) as well as more than 100,000 disclosed compounds containing the fragment. More than 90 kinases have been targeted by molecules containing 7-azaindole, and the paper provides a list of 70 PDB structures of 37 different kinases bound to molecules containing the moiety.

Finally, in J. Med. Chem., Brian Raymer and Samit Bhattacharya (Pfizer) survey the universe of “lead-like” drugs. Among the most highly prescribed small molecule drugs, 36% have molecular weights below 300 Da. Only 28 of 174 drugs approved between 2011 and 2017 fall into this category, consistent with the increasing size of newer drugs. The researchers discuss 16 recently approved drugs, and find that 13 have very high ligand efficiencies (at least 0.4 kcal mol^-1 per heavy atom). As noted above, optimization often entails adding molecular weight by growing or linking, and the researchers suggest that alternative strategies such as conformational restriction and truncation also be investigated.

And with that, Practical Fragments wishes you a happy new year. Thanks for reading some of our 686 posts over the past decade plus, and please keep the comments coming!

Fragments in the clinic: eFT508

Multiple clinical candidates derived from fragments were described at the recent CHI FBDD Meeting. The story behind one of these has just appeared in J. Med. Chem. in a paper published by Siegfried Reich and colleagues at eFFECTOR Therapeutics.

The researchers were interested in mitogen-activated protein kinase interacting kinases 1 and 2 (MNK1/2), which appear to be important in tumorigenesis but not normal cells. As Paul Sprengeler described it, they started with a “library” of just six fragments – four from the literature and two designed. (The company began in a law office, so hands-on experiments were initially limited.) Some might ask whether this constitutes FBDD, but in the end it’s not the size of your library that counts, but what you do with it.

All the fragments had good affinity, and the researchers were able to obtain crystal structures of four of them bound to MNK2. Optimization proceeded on all six of the fragments, but compound 1 was considered particularly attractive due to its high ligand efficiency and multiple vectors for growing.

Initially the researchers deconstructed the bicyclic ring to compound 7, which led to a 10-fold loss in potency but reduced the molecular weight and lipophilicity. As they note, “loss of potency in exchange for improved physicochemical properties is an often overlooked yet powerful optimization strategy in medicinal chemistry.” Too often people focus on binding over drug-like properties, so it is refreshing to see smart tradeoffs explicitly acknowledged.

Next, the researchers cyclized the molecule to form a lactam and remove one hydrogen bond donor. This also improved the affinity (compound 10). Replacing the phenyl ring in compound 10 with a pyridone in compound 12 further reduced the lipophilicity and improved the selectivity due to non-covalent interactions between a non-conserved cysteine residue and the heterocyclic ring. More optimization led to eFT508.

In addition to low nanomolar potency against both MNK1 and MNK2 in biochemical and cell-based assays, eFT508 is metabolically stable, permeable, and orally bioavailable in mice, rats, dogs and monkeys. The molecule showed good activity in several mouse xenograft models, and tissue samples revealed reduced phosphorylation of the substrate protein eIF4E, as expected. Unlike the MTH1 story last week, in which a selective chemical probe devalidated the target, the results with eFT508 suggest that inhibiting MNK1 and MNK2 has merit, and the compound is currently in four clinical trials for both solid tumors and lymphoma.

Like the story last week, this program progressed rapidly: just 110 compounds, 30 crystal structures, and 1 year to the development candidate. This turned out to be somewhat lucky, as it took another two years to find an equivalently attractive backup candidate. It is also an excellent example of how a non-selective fragment (a purine, found in ATP!) can be turned into a selective molecule. And finally, this is another nice example of how even a public, generic fragment can lead to an attractive chemical series. There are myriad published fragments bound to legions of targets, and it’s worth keeping these in mind whether or not you have in-house biophysical screening capabilities.

Fragment optimization without purification

Compound purification can be a major hassle: separating the desired product from starting materials, reagents, and byproducts often takes far longer than making the compound in the first place. As we’ve previously noted, this is especially true for small, polar fragments – which are particularly attractive for drugs. Two new papers address this challenge. (Shameless plug: my company Carmot Therapeutics also has a solution to this problem.)

In J. Med. Chem., Paul Brough and Vernalis colleagues describe their discovery of inhibitors of all four isoforms of pyruvate dehydrogenase kinase (PDHK), potential targets for diabetes and oncology. The ATP-binding site of these four enzymes is similar to that of oncology target HSP90, in which Vernalis has a long-standing interest.

A screen of 1063 fragments (each at 0.5 mM) against PDHK-2 yielded 78 hits that were positive in three different NMR-based assays and also ATP-competitive. These yielded a whopping 43 structures when soaked into crystals of the related isoform PDHK-3. Compound 6 was one, and the binding mode was very similar to that previously seen for the same fragment with HSP90. Fragment growing rapidly led to molecules such as compound 8, with low micromolar potency. This compound was almost equipotent against HSP90, but modeling suggested that it might be possible to further grow this molecule in a direction that would be accommodated in the PDHKs but not in HSP90.

The next step was to make a bunch of analogs, and here's where avoiding purification becomes advantageous. Specifically, the researchers turned to off-rate screening (ORS), which entails making compounds and then testing the impure mixtures using surface plasmon resonance (SPR) to look for those which dissociate more slowly. Since off-rate is not dependent on the concentration of ligand, a low yield shouldn’t change the results of the assay.

An initial library of 56 compounds led to the discovery of compound 18, and subsequent libraries and medicinal chemistry ultimately yielded VER-246608, which is a potent pan-PDHK inhibitor. As designed, it is also completely inactive against HSP90. The molecule is described more thoroughly in this Oncotarget paper, which reveals that despite activity against PDHKs in cells, VER-246608 is not particularly effective at slowing the proliferation of cancer cells. Still, it does appear to be a useful chemical probe for further exploring the biology of the PDHKs.

Shifting methods but staying with the theme of assaying impure compounds brings us to a paper in SLAS Discovery by Sten Ohlson, Brian Dymock, and colleagues at Nanyang Technological University and the National University of Singapore. The protein tested was HSP90, and the method used was weak affinity chromatography, or WAC (see here, here, and here).

Like SPR, WAC also uses an immobilized protein. However, whereas SPR provides the (kinetic) off-rate, WAC provides the (thermodynamic) dissociation constant, which is calculated from the change in retention time of the molecule as it passes through a column containing protein-bound resin. In this case the researchers synthesized a mixture of five different compounds which varied from 7-24% of the mixture. This crude sample was analyzed by WAC, and the resulting dissociation constants, ranging from 48-147 µM, were satisfactorily similar to the values obtained using pure compounds.

Both of these approaches should accelerate screening and facilitate the analysis of complicated mixtures, such as natural product extracts. It will be fun to watch for more examples.

Lead Generation: Methods, Strategies, and Case Studies

Lead generation refers to that point in drug discovery when initial screening hits against a target are wrought into compelling chemical matter. This chemical matter is often plagued with deficiencies in terms of potency, pharmacokinetics, or novelty, yet it provides a starting point for further optimization. This is the subject of a massive (800+ pages!) new two-volume work edited by Jörg Holenz (GlaxoSmithKline, formerly AstraZeneca) as part of Wiley’s Methods and Principles in Medicinal Chemistry series. Readers of this blog will not be surprised to find that fragments play a major role; indeed, the molecule on the cover of the book came out of FBLD. I won’t attempt to summarize all 25 chapters here, but will simply highlight those most relevant to FBLD.

Mike Hann (GlaxoSmithKline) sets the stage in chapter 1 by briefly describing the characteristics of successful leads. He emphasizes the importance of physicochemical properties and avoiding molecular obesity, and how judicious use of metrics can help navigate away from perilous chemical space. He also summarizes internal programs that again demonstrate that fragment-derived leads tend to be smaller and less lipophilic than those from other lead discovery techniques.

In chapter 3, Udo Bauer (AstraZeneca) and Alex Breeze (University of Leeds) discuss the concept of ligandability – the ability of a target to bind to a small molecule with high affinity. Fragments are ideally suited for assessing ligandability, and the researchers briefly describe fragment-based experimental and computational approaches to do so. They also include a nice 11-point summary of factors to consider when starting lead generation on a new target, ranging from the presence of small-molecule binding sites to the number of patent applications.

Chapter 6, by Ivan Efremov (Pfizer) and me, is entirely about fragment-based lead generation. I'm undoubtedly biased, but I think it provides a self-contained and fairly detailed guide to FBLD, including topics such as screening methods, hit validation, metrics, hit optimization, fragment growing vs fragment linking, and case studies on vemurafenib, BACE, MMP-2, LDHA, venetoclax, MCL-1, and GPCRs.

Helmut Buschmann and colleagues at RD&C Research, Development, and Consulting, focus in chapter 9 on optimizing side effects of known molecules to develop new drugs, but they also discuss some interesting older work reporting that 418 of 1386 drugs contain other drugs as internal fragments.

Chapter 12, by Dean Brown (AstraZeneca), is devoted to the hit-to-lead stage, and much of his advice is applicable to FBLD. Dean also includes a fantastic metaphor to illustrate the size of chemical space: "if a typical corporate screening collection were to fit on a postcard, the rest of the earth is the amount of available drug-like space." This assumes a million-compound library and a conservative estimate of 10²³ drug-sized molecules, so if anything it is an understatement.

Molecular recognition is critical for both FBLD and lead generation in general, and this is the topic Thorsten Nowak (C4X Discovery Holdings) tackles in chapter 13. He covers key areas such as thermodynamics, emphasizing the importance of enthalpy while acknowledging the difficulty of prospectively using thermodynamic data. The role of water and halogen bonds are covered, along with some freakishly high ligand efficiency values. There are a couple errors: one paper is categorized as using dynamic combinatorial chemistry when in fact it actually used static libraries, and Tethering is confused with Chemotype Evolution, but overall there's lots of good stuff here.

Biophysical methods are covered in chapter 14, by Stefan Geschwindner (AstraZeneca). These include NMR, SPR, ITC, thermal shift assays, native mass spectrometry, microscale thermophoresis, and more.

Chapter 16, by Ken Page and colleagues at AstraZeneca, discusses "lead quality." This often entails various metrics, from simple ones such as ligand efficiency and LLE to more complicated attempts to predict clinical dosages. Although it is easy to poke fun at metrics, most thoughtful scientists find them useful for making sense of the reams of data generated in lead optimization campaigns.

Chapter 17, by Steven Wesolowski and Dean Brown (both AstraZeneca), is arguably the most entertaining. Entitled "The strategies and politics of successful design, make, test, and analyze (DMTA) cycles in lead generation," it is replete with pithy quotes and even an original (and highly geeky) cartoon. Along with multiple examples, the chapter formulates plenty of questions to consider during lead optimization, and ends with a particularly relevant quote by Billings Learned Hand: “Life is made up of a series of judgments on insufficient data, and if we waited to run down all our doubts, it would flow past us.”

In chapter 23, Sven Ruf and colleagues at Sanofi-Aventis Deutschland describe a success story generating leads against cathepsin A, a target for cardiovascular disease. HTS yielded three different chemical series with sub-micromolar activities, each with different liabilities. Crystallography revealed their binding modes, and this allowed the team to mix and match fragments across the different series to generate a molecule that ultimately went into the clinic. Although this may not be classic FBLD, it does seem to be a good case of using concepts from the field, or fragment-assisted drug discovery.

A similar, if less directed, approach is the subject of chapter 25, the last in the book. Pravin Iyer and Manoranjan Panda (both AstraZeneca) describe "fragmentation enumeration," in which known drugs or clinical candidates are fragmented into component fragments and recombined. On some level the fragments themselves are likely to be privileged; the researchers cite the famous quote by Sir James Black that "the most fruitful basis of the discovery of a new drug is to start with an old drug." Most of the work is computational, although one molecule derived from the approach has encouraging cellular activity against Mycobacterium tuberculosis.

There's far more to this book than could be listed even in this relatively long post, including multiple case studies, so for those of you who are interested in lead generation definitely check it out!

Off-rate screening (ORS)

Molecules that dissociate slowly from their target proteins are potentially useful because they can have a long-lasting effect even if they are rapidly cleared from circulation. However, it is next to impossible to predict whether a molecule will dissociate slowly or not. Moreover, the correlation with binding affinity is poor: weak binders generally don’t stay bound to their target for long, but even tight binders often rapidly dissociate. In the early stages of lead discovery most folk are focused on affinity, and it is usually only much later that kinetics enters in. In a new paper in J. Med. Chem., James Murray, Paul Brough, and colleagues at Vernalis introduce a technique that moves kinetics to the front of the line.

The technique, off-rate screening (ORS), relies on surface plasmon resonance (SPR), which is already commonly used to study binding kinetics. The trick here is using SPR to screen products in unpurified reaction mixtures. An initial fragment with known affinity is modified, and products screened for slower dissociation. Of course, the concentration of desired compound is likely to vary from mixture to mixture, but the great thing about looking at compound dissociation is that it is a zero order reaction: it does not depend on concentration. The researchers use mathematical simulations to show that even if the yield is only 5%, a product with a 10-fold slower dissociation rate constant could still be detected. Since off-rates can vary by orders of magnitude, this is not such a high bar.

Of course, simulations are one thing, but how does the technique actually work in practice? The researchers show examples on two targets, one using some of the early compounds for their HSP90 program, the other some of their PIN1 inhibitors. For PIN1, the researchers resynthesized some of the molecules in plastic tubes, which caused leaching of plastic into the reaction mixtures. Nonetheless, for both proteins the dissociation rate constants measured for unpurified reactions were very close to purified molecules, generally differing by less than 30%.

The researchers also tried subjecting compounds to eleven reaction conditions typically used in medicinal chemistry, evaporating the solvent, and testing the products; the idea was to see if the reagents or other components in the reaction mixture would interfere with the assay. Happily in all cases the dissociation rate constants differed by less than 20%, again pointing to the robustness of ORS.

Of course, as with any technique, there are limitations. Since the screening compounds are not purified from their starting materials, the desired products must dissociate sufficiently slowly from the protein to be distinguishable from other components in the reaction mixture; dissociation rate constants greater than about 1.2 s^-1 appear to be challenging. Also, if the starting material itself has a slow dissociation rate from the protein, it may be difficult to differentiate this from a low yield of slowly dissociating product. The researchers note that both cases could be addressed by changing the temperature, either lowering it to slow the dissociation rate constant or raising it to increase it.

All in all this is a nice approach, and it will be interesting to see how widely it catches on.

Leave Them Asking for More

ret·ro·spec·tive  (rtr-spktv) adj.

1. Looking back on, contemplating, or directed to the past.

2. Looking or directed backward.

3. Applying to or influencing the past; retroactive.

I would add: 4. Looking back on the past, to influence the future.

In this vein, a recent paper by Ferenczy and Keserű in J. Med Chem looks back on hit-lead optimizations derived from fragment starting points.  In this very interesting paper, they look at 145 fragment programs and evaluate the properties of the original hit and then again as it progresses into the lead.  Of the 145 programs, these were aimed at 83 proteins of which 76 are enzymes, 6 are receptors, and 1 is an ion channel.  These programs evolved into leads, tools, and clinical candidates.  The authors set out to answer three questions: 1. do fragments eliminate the risk of property inflation, 2. how do ligand efficiency metrics support fragment optimizations, and 3. what is the impact of detection method, optimization strategy, and company size on the optimization.

Table 2 shows the median the calculated properties for the hits and optimized compounds.  The pIC50 improved by roughly three orders of magnitude, but ligand efficiency (LE) stayed roughly the same.  Log P increased but SILE did not.  SILE was a metric I was not familiar with and is calculated by pIC50/(HAC)^0.3.  SILE is a size-independent metric of ligand efficiency.  I won't attempt to reproduce all the graphs they generated; get the paper.  Some interesting data points: the median fragment hit had 15 heavy atoms (see related poll here), the median size of leads is 28 heavy atoms, and good fraction of hit-lead pairs changed less than 5 heavy atoms (which of course is well known from here). So, what is the answer for their first question?  If you are looking at something like LE, then these hit-lead pairs maintain the efficiency.  If you are looking at logP, then the answer is no, the hit-lead pairs get greasier.  I would really like to see more granularity here (see What's Missing below).  Their major comment here is that SILE and LELP (work by the author's previously reviewed by Dan) are the two best metrics to monitor as hit-lead optimization is underway.  Increases in both metrics correlate with increases in FBDD programs. 

 

They then looked at the  screening method.  The breakout of primary screening (in their definition the first one listed when multiple methods were used) was 38% biochemical, 25% NMR, 18% X-ray, and 11% virtual.  This is an interesting contrast to these results; SPR is not the dominant screening technique (8% tied with MS).  So, does this mean, >40% of practitioners are using SPR, but as a secondary screen?

Table 4 then does a pairwise comparison of the metrics based upon primary screen origin of the fragment (see What's Missing below).  Biochemical screens yield the most potent hits (4.75) while NMR (3.53) has the least potent.  X-ray has the smallest hits( 13 heavy atoms) while virtual screening the largest (17 HA).  I don't think any of this is surprising; the authors point out that hit properties exhibit a significant dependence on the method used. It is noted that optimization tends to diminish differences in hit properties.  Again, I think this is not surprising; thermodynamics and medchem are all the same no matter how big or small the molecule.  They do point out that biochemical based hits preserve their advantage after optimization, primarily relative to NMR.  They posit that the difference is that more potent compounds need less "stuff" to become sufficiently potent, and thus have a better mix of medchem for potency and other property optimization.  Weaker starting points need more bulk, more atoms, to become equipotent with biochemical starting points, they suggest.  Lastly, they show that structural-based optimization efforts are better than those without structural information.  The structural information comes from X-ray primarily (52%) and NMR (10%).  Interestingly, they reference this poll from the blog on whether you need structural information to prosecute fragments.  [As an aside, since I wrote that blog post and it was reference in a paper, do I get to add it to my resume?]

Finally, they break the originating labs into three categories: academic (18%), small/medium enterprises (SME)( 37%), and Big Pharma (45%).  The SME results are superior to those achieved by the academics and Big Pharma.  Their explanation is that SME's tend to be more platform focused and predominantly ensure "structural" enablement of targets.  75% (40/53) optimizations at SMEs used protein structural information, while "only" 62% of those at Big Pharma did.  They do not rule out the differences in target selection at those two different groups of companies.  I would like to propose an alternative hypothesis (tongue-not-entirely-in-cheek): Big Pharma has "old crusty" chemists who don't understand fragments and thus just glom on hydrophobic stuff to increase potency because "that's how we always do it" while SME have innovative chemists.  And of course academia is just making tool compounds and crap.

One thing that I would like to emphasize is that Ro3, metrics, and so on should not be used as hard cutoffs.  As shown in Figure 11, even compounds that are outside the "preferred" space can reach the clinic.  The best way to view them is akin to the Pirate Code; they are more guidelines than rules. 

 

What is Missing?  The supplemental information (which the authors are willing to kindly share) does not break out the targets into specific classes.  However, they do list each target, so it should be easy to add this a data column.  More importantly, they do not break out hit-lead pairs into those that were optimized for use as tools, clinical candidates, and leads (and which of the leads died).  Tools are never supposed to look like leads (but you are lucky if they do), so their inclusion here can be biasing the results.  Although, it is likely that the of the 145 not very many of the examples are strictly tools; it would be nice to know though.   

I am struck by the information denseness of Table 4 and wish that instead of pairwise comparisons, they had instead simply list the hit-lead metrics for each methodology.  I think there is gold to be mined in Table 4 and just like real gold is hard to find.

 

I have not addressed every single point made by the authors.  I, for one, am hoping that they will continue their analyses (especially with an eye to some of What is Missing).  I hope that there will be a significant amount of discussion around these points.  I will make sure we hit on this at the breakfast roundtable at the upcoming CHI FBDD event in SD (I even have the same pithy title as last year!)

Poll Results--HAC vs. MW

Our poll asking what denominator people use for the ligand efficiency metrics.  This idea for this poll came from these posts at In The Pipeline.  Of the 38 respondents, 8 "don't need no steenkin' metrics".  Of the remaining 30 answers, 27 people use heavy atom count, 1 uses both, and only 2 use molecular weight. 

So, in terms of everyday atoms, I think people can agree that heavy atom count makes the most sense.  But, Derek is still trying to figure out what to do with heavy halogens.  Do halogens need to be treated differently?  My thoughts are that they don't in the hit generation (HG) stage, but in lead optimization stage they would.   I have always thought that ligand metrics are most germane to the HG stage and less useful once you are trying to optimize things like PK/PD properties.  If a heavy halogen, makes it from hit confirmation, hit expansion, and into lead optimization, it is most likely doing something, so why penalize it?  

Am I thinking about this too naively?