18 January 2021

Does configurational entropy explain why fragment linking is so hard?

Linking two weak fragments to get a potent binder is something many of us hope for. Unfortunately, as a poll taken a few years back indicates, it often doesn’t work. But why? This is the question tackled by Lingle Wang and collaborators at Schrödinger and D. E. Shaw in a recent J. Chem. Theory Comput. paper.
When a ligand binds to a protein it pays a thermodynamic cost in terms of lost translational and orientational entropy. By linking two fragments, this cost is paid only once instead of twice. In theory this should lead to an improvement of 3.5-4.8 kcal/mol in binding energy, resulting in a 400-3000-fold improvement in affinity over what would be expected from simple additivity. As we noted here, this is possible, though rare. Linker strain often takes the blame as a primary villain. But is there more to the story?
The researchers computationally examined published examples of fragment linking (most of which we’ve covered on Practical Fragments) using free energy perturbation (FEP) to try to understand why the linked molecules bound more or less tightly than expected. Impressively, they were able to computationally reproduce experimentally derived numbers, and by building a thermodynamic cycle they could extract the various components of the “connection Gibbs free energy.” These included changes in binding mode or tautomerization, linker strain or linker interactions with the protein, and the previously mentioned entropic benefits of fragment linking.
The analysis also identified two additional components. If two fragments favorably interact with each other, covalently linking them may not give as much of a boost. This concept had been considered decades ago, though the current work provides a more general understanding.
The more important factor appears to be what the researchers refer to as “configurational entropy.” The notion is that even when a fragment is bound to a protein, both the ligand and protein retain considerable flexibility, which is entropically favorable. Linking two fragments reduces the configurational entropy of each component fragment, and the linked molecule binds less tightly than would be expected. The researchers argue that this previously unrecognized “unfavorable change in the relative configurational entropy of two fragments in the protein pocket upon linkage is the primary reason most fragment linking strategies fail.” They advise that maintaining a bit of flexibility in the linker can help, as has been previously suggested.
This is an interesting analysis, and explicitly considering configurational entropy is likely to improve our understanding of molecular interactions. But is it really the main barrier to successful fragment linking? The researchers explore only nine different protein-ligand systems, though they did consider multiple linked molecules for three of these (pantothenate synthetase, RPA, and LDHA). Still, these represent just a fraction of the 45 examples collected in a recent review, and they only considered one somewhat contrived case (avidin) in which especially strong superadditivity was observed. It will be interesting to see whether the analysis holds true for more examples of fragment linking.

11 January 2021

Hundreds of fragments hits for the SARS-CoV-2 Nsp3 Macrodomain

COVID-19 will be with us for some time. Despite the unprecedented speed of vaccine development, it is worth remembering that humanity has only truly eradicated two widespread viral diseases, smallpox and rinderpest. Thus, the long march of small molecule drug discovery against SARS-CoV-2 is justified. In a paper recently posted on bioRxiv, Ivan Ahel and more than 50 multinational collaborators take the first steps.
Last year we highlighted two independent crystallographic screens against the main protease of SARS-CoV-2. Another potential viral target is the macrodomain (Mac1) portion of non-structural protein 3 (Nsp3), an enzyme which clips ADP-ribose from modified proteins, thus helping the virus evade the immune response.
The researchers soaked crystals of Mac1 against a total of 2683 fragments curated from several collections. This yielded 214 hits, and most of the structures were solved at high resolution (better than 1.35 Å). About 80% of the fragments bound in the active site, with many binding in the adenosine sub-pocket. Two different crystal forms were used for soaking, and one set of 320 fragments was soaked against both. Interestingly, this yielded a hit rate of 21% for one crystal form and just 1.3% for the other. Even more surprising, of the five hits found in both crystal forms, only two bound in the same manner in both. This is a clear demonstration that it is worth investing up-front effort to develop a suitable crystal form of a protein before rushing into soaking experiments.
Independently, the researchers computationally screened more than 20 million fragments (mostly from ZINC15) against the protein using DOCK3.7, a process which took just under 5 hours on a 500-core computer cluster. Of 60 top hits chosen for crystallographic soaking, 20 yielded structures, all at high resolution (0.94-1.01 Å). The ultra-high resolution structures revealed that four fragments had misassigned structures (wrong isomers), which long-time readers may not find surprising. Importantly, most of the 20 experimentally determined structures confirmed the docking predictions.
A strength and weakness of crystallographic screening is that it can find extraordinarily weak binders, which may be difficult to optimize. To see whether they could independently verify binding, the researchers tested 54 of the docking hits in a differential scanning fluorimetry (DSF) assay. Ten increased thermal stability, and all of these had yielded crystal structures. Only four of 19 fragments tested yielded reliable data in isothermal titration calorimetry (ITC) assays, but encouragingly these four also gave among the most significant thermal shifts in the DSF assay. Finally, 57 of the docking hits and 18 of the crystallographic hits were tested in a homogenous time-resolved fluorescence (HTRF) based peptide-displacement assay, yielding 8 and 3 hits respectively, the best with an IC50 of 180 µM.
This paper is a tour de force, and may represent the largest collection of high-resolution crystallographic fragment hits against any target. Laudably, all 234 of the crystal structures have been released in the public domain, and the researchers have already suggested ideas for merging and linking. As they point out, many of the fragments bind in the adenine pocket, so selectivity will be an issue not just against human macrodomains but also against kinases and other ATP-dependent enzymes. Still, as the dozens of approved kinases inhibitors demonstrate, achieving selectivity is possible. 
From a technology perspective, this publication affirms the rising power of both crystallographic and computational screening. Indeed, the hundreds of crystal structures will themselves be useful input for training new computational methods. And from a drug discovery perspective, each of these fragments represents a potential starting point for SARS-CoV-2 leads.
Let’s get busy!

04 January 2021

Fragment events in 2021

Gotten vaccinated yet? Don't worry - the first few conferences of the year will be virtual, but hopefully we'll be meeting in person later this year.

March 9-12:  While not exclusively fragment-focused, the Second NovAliX Virtual Conference on Biophysics in Drug Discovery will have several relevant talks. You can read my impressions of the 2018 event here, the 2017 Strasbourg event here, and Teddy's impressions of the 2013 event herehere, and here.
May 18-19: CHI’s Sixteenth Annual Fragment-Based Drug Discovery, the longest-running fragment event, will again be held virtually. This is part of the larger Drug Discovery Chemistry meeting, running May 18-20. You can read impressions of the 2020 virtual meeting here, the 2019 meeting here, the 2018 meeting here, the 2017 meeting here, the 2016 meeting here; the 2015 meeting herehere, and here; the 2014 meeting here and here; the 2013 meeting here and here; the 2012 meeting here; the 2011 meeting here; and 2010 here.
September 27-30: CHI’s Nineteenth Annual Discovery on Target returns to the real world - or at least Boston. As the name implies this event is more target-focused than chemistry-focused, but there are always plenty of FBDD-related talks. You can read my impressions of the 2020 virtual event here, the 2019 event here, and the 2018 event here.

December 16-21: What better place to say goodbye to COVID than Hawaii? Postponed from last year, the second Pacifichem Symposium devoted to fragments will be held in Honolulu. Pacifichem conferences are normally held every 5 years and are designed to bring together scientists from Pacific Rim countries including Australia, Canada, China, Japan, Korea, New Zealand, and the US. Here are my impressions of the 2015 event.
Know of anything else? Please leave a comment or drop me a note!

28 December 2020

Review of 2020 reviews

An old curse runs, "may you live in interesting times." And 2020 has been interesting indeed. Amid all the tumult, Practical Fragments will maintain its tradition of ending the year with a post highlighting conferences and reviews.
Despite the travel restrictions caused by COVID-19, some conferences did go ahead, adapted to online formats: I highlighted CHI’s Fifteenth Annual Fragment-based Drug Discovery and their Eighteenth Annual Discovery on Target. Although these were quite successful, I think most of us are looking forward to returning to in-person events sometime in the coming year.
Perhaps because so many people were stuck working from home, the number of reviews of potential interest to fragment fans has soared to a record number of more than twenty. I’ve tried to group these thematically.
If you’re looking for a concise yet thorough review, Harren Jhoti and colleagues at Astex provide one in Biochem. Soc. Trans. Harren is one of the pioneers of FBDD, and the review touches on library design, detection of fragment binding, and fragment to lead strategies. A review in Front. Mol. Biosci. by Qingxin Li (Guangzhou Sugarcane Industry Research Institute) goes into more detail on fragment screening, optimization, and biological targets.
For the past five years a few fragment fanciers (myself included) have been writing annual reviews in J. Med. Chem. covering fragment-to-lead success stories from the previous year, each with a handy table showing fragment, lead, and key parameters. The 2018 edition, led by yours truly (Frontier Medicines), was published at the beginning of the year, while the 2019 edition, led by Wolfgang Jahnke (Novartis), just came out a few weeks ago. At the risk of self-promotion, both are well worth perusing to see the growing diversity of targets and emerging trends, such as covalent fragments.
Biophysical methods are by far the most commonly used for finding fragments, and an excellent overview of thermal shift, SPR, and NMR by Joe Coyle and Reto Walser (Astex) appears in SLAS Discovery. The goal is “to help the anxious biophysicist withstand the relentless unforeseen,” and the paper provides loads of practical advice. For example, over more than 50 thermal shift screens, “we have never derived anything useful from negative Tm shifts.” The researchers note that “SPR is particularly user-friendly and particularly prone to artifact, overinterpretation, and varying degrees of frustration.” As for validating ligand-observed NMR hits crystallographically, rates range from 5% to 80%.
As we noted earlier this year, crystallography is becoming increasingly dominant in fragment screening, and in Molecules Laurent Maveyraud and Lionel Mourey (Université de Toulouse) provide an overview of the process, covering theory, workflow, practical aspects, pitfalls, examples, and other emerging methods. David Stuart and colleagues at Diamond Light Source discuss structural efforts on SARS-CoV-2 proteins in an open-access paper in Biochem. Biophys. Res. Commun. As of late October this included more than 500 released structures of 16 different proteins. Efforts against the main protease (which I reviewed in Nat. Commun.) have led to molecules with mid-nanomolar activity, and the researchers rightly highlight the worldwide collaboration that has led to such rapid progress.
NMR is of course a biophysical technique, but there are so many papers this year that it makes sense to group them into their own section. Ray Norton (Monash Institute of Pharmaceutical Sciences) and Wolfgang Jahnke (Novartis) introduce a special issue of J. Biomol. NMR focused on “NMR in pharmaceutical discovery and development” by briefly summarizing the state of the art and introducing 13 articles, one of which we covered previously and three of which are highlighted below.
“NMR in target driven drug discovery, why not?” ask Gregg Siegal and collaborators at ZoBio and Gotham in an (open access) J. Biomol. NMR review. In addition to characterizing small molecules, proteins, and their interactions, the researchers present cases studies in which NMR data has helped clarify a crystallographic protein-ligand structure, or even suggested that the crystal structure represented at most a minor conformation in solution.
In other words, NMR is “the swiss army knife of drug discovery,” as Reto Horst and colleagues at Pfizer put it in another J. Biomol. NMR review. The researchers describe successful NMR fragment screens against difficult targets such as an ion channel and a large (145 kDa) trimeric enzyme. They also make a good case for using NMR to determine the solution conformations of small molecules early in a project, a strategy that has paid off in more than 15 Pfizer projects over the past six years.
Benjamin Diethelm-Varela (University of Maryland) focuses on using NMR for “fragment-based drug discovery of small-molecule anti-cancer targeted therapies” in ChemMedChem. This is a thorough yet accessible overview of FBDD, ligand- and protein-based NMR methods, plus ten case studies. “A practical perspective on the roles of solution NMR spectroscopy in drug discovery” is provided by Qinxin Li and CongBao Kang (A*STAR) in Molecules. As the title suggests, this review is fairly broad, and includes an interesting section on NMR screening in cells.
All these papers might have you thinking that NMR is a “Gold Standard,” and that phrase does indeed appear in the title of another Molecules review by Abdul-Hamid Emwas (King Abdullah University of Science and Technology) and a multinational group of collaborators. This is a large (66 page) monograph with 455 references and is particularly detailed on various NMR techniques; if you want to see the pulse sequence of the HSQC experiment or review the Einstein-Stokes equation this is the place to turn.
In addition to the six reviews on NMR above, two specifically cover 19F NMR. The first, from the J. Biomol. NMR special issue by Claudio Dalvit (Lavis) and colleagues, focuses on fluorine NMR functional screening, or n-FABS. This paper provides an excellent theoretical and practical overview of the technique, and includes a handy table of 17 published case studies. And in Prog. Nuc. Mag. Res. Spect. Peter Howe (Syngenta) reviews “recent developments in the use of fluorine NMR in synthesis and characterization.” As the title suggests, much ground is covered, from spectrometer technology to quantum chemistry calculations, and there is a short section on fragment-based screening.
Turning to in silico techniques, Floriano Paes Silva Jr. and collaborators at LaBECFar and several other (mostly) Brazilian institutes provide an open-access overview in Front. Chem. After summarizing FBDD they describe how computational techniques can help along the way, from druggability prediction to docking, de novo design, and assessment of ADMET properties and synthetic accessibility. The review ends with several case studies.
In an open-access article in Drug Disc. Today, Stefano Moro and colleagues at University of Padova focus on “the rise of molecular simulations in fragment-based drug design.” This accessible overview covers hotspot identification, hit identification and characterization, and hit to lead optimization, and includes a nice section on free energy perturbation.
Other topics
Molecular properties are critical for developing good drugs, and in J. Med. Chem. Christopher Tinworth (GlaxoSmithKline) and Robert Young (Blue Burgundy) “appraise the rule of 5 with measured physicochemical data.” This is packed full of good stuff including a supplementary table with calculated and measured data for hundreds of compounds. The summary is that molecular weight is much less important than (measured) lipophilicity and hydrogen bond donors. “Good practice is all about compromise, aiming to maximize efficacy and efficiency while navigating many potential pitfalls in molecular optimization.” People sometimes obsess over rules vs guidelines, and the researchers close by stating that “rules are for the obedience of fools and guidance of the wise.”
As a poll from several years ago suggested, fragment linking tends to be less common than fragment growing, though it can work spectacularly. In J. Med. Chem. Isabelle Krimm and colleagues mostly at Université de Lyon review 45 successful fragment linking case studies (though it would have been appropriate for them to acknowledge Practical Fragments for the clearly borrowed table of clinical compounds). While by no means exhaustive, this is a useful resource. Interestingly, only 20% of the examples display superadditivity.
Target-guided synthesis (TGS) can be thought of as a special case of fragment linking. In J. Med. Chem., Rebecca Deprez-Poulain and colleagues at Université de Lille review kinetic TGS, in which two components react irreversibly with one another in the context of a protein to form a higher-affinity binder. Kinetic TGS may have some practical advantages over reversible TGS (or dynamic combinatorial chemistry), but as the researchers note most examples start with compounds larger than fragments, and thus only 38% of examples lead to products with a molecular weight less than 500 Da. This could partly explain why only 6 of the 50 reported examples have gone into animal studies.
Finally, György Keserű and collaborators at the Hungarian Research Centre for Natural Sciences review covalent fragment-based drug discovery in Drug Discovery Today (open access). Library design and validation is well-covered, as are various methods for screening covalent fragments, and there is a handy table of some four-dozen published examples. Given the increasing popularity of covalent FBLD, this contribution should be of wide interest.
When I wrote my concluding post for 2019, COVID-19 was an obscure and nameless disease, and SARS-CoV-2 had not even been identified. I ended with, "may 2020 bring wisdom, and progress." We've gained both, though the cost has been incalculable. So I'll just close this post by thanking you for reading and commenting.

14 December 2020

Benchmarking docking methods: a new public resource

Despite advances in crystallography, obtaining structures of fragments bound to proteins is still often elusive. Computational docking is likely to forever be faster than experimental methods, but how good is it? A new paper in J. Chem. Inf. Mod. by Laura Chachulski (Jacobs University Bremen) and Björn Windshügel (Universität Hamburg) assess four popular methods and also provide a public validation set for others to use.
When evaluating fragment docking methods, it is essential to have a well-curated set of experimental structures. To this end, the researchers started by combing the PDB for high quality, high resolution (< 2 Å) structures of protein-fragment complexes. They used automated methods to remove structures with poor electron density, close contacts with other ligands, and various other complications. Further manual curation yielded 93 protein-ligand complex structures. The fragments span a relaxed rule-of-three, with 7 to 22 non-hydrogen atoms (averaging 13) and ClogP ranging from -4.1to 3.5 (averaging 1.1). I confess that some choices are rather odd, including oxidized dithiothreitol, benzaldehyde, and γ-aminobutyric acid. The researchers might have saved themselves some effort, and obtained a more pharmaceutically attractive set, by starting with previous efforts such as this one.
Having built their benchmark data set, called LEADS-FRAG, the researchers next tested AutoDock, AutoDock Vina, FlexX, and GOLD to see how well they would be able to recapitulate reality. The results? Let’s just say that crystallographers look likely to have job security for some time.
Only 13 of the 93 protein-fragment complexes were correctly reproduced as the top hit using all four methods (even with a reasonably generous RMSD cutoff criterion of < 1.5 Å).There were 18 complexes that none of the methods predicted successfully. Across the four methods, the top-ranked poses were “correct” 33-54% of the time. Docking methods usually provide multiple different poses with different scores; up to 30 were considered here. Looking at lower-ranked poses increased the number of successes to 27 of the 93 fragments, while only three failed in all methods. Overall, the correct structure was present among the poses in 53-86% of cases. Changing the scoring function sometimes led to further improvements.
Why were some fragments more successfully docked than others? Fragments that were more buried within the protein (lower solvent-accessible surface area, or SASA) yielded better predictions than those that were more solvent-exposed. The researchers did not report on the effect of rotatable bonds; intuitively, one might think that a more flexible fragment would be harder to dock. A study we highlighted nearly ten years ago found that fragments with higher ligand efficiency also had higher docking scores, and it would be interesting to know if that reproduced with this larger data set.
The researchers conclude by noting that “these programs do not represent the optimal solution for fragment docking.” I think this is a fair assessment. And as the researchers acknowledge, the bar was set low: compounds were docked against the crystal structure of the protein with ligand computationally removed. In the real world, proteins often change conformation upon ligand-binding, which would make docking even more difficult.
In addition to trying to determine how a specific fragment binds, it can also be valuable to computationally screen large numbers of fragments. The programs used here took between 10 seconds and 42 minutes per ligand, but as we highlighted last year speed continues to increase.
Most importantly, the public availability of LEADS-FRAG will allow others to assess their own computational approaches. It will be fun to revisit this topic in a few years to see how much things have improved.

07 December 2020

Fragments vs LpxC, two ways

Gram negative bacteria such as Pseudomonas aeruginosa are a continuing threat, and antibacterial drug discovery is not keeping pace. The enzyme UDP-3-O-acyl-N-acetylglucosamine deacetylase (LpxC) is critical for the synthesis of the bacterial cell wall lipopolysaccharide. In a new J. Med. Chem. paper, Yousuke Yamada, Rod Hubbard, and collaborators at Taisho and Vernalis describe progress against this target.
LpxC is a zinc hydrolase, and although previous potent inhibitors have been reported against the metalloenzyme, these contained hydroxamate moieties. Unfortunately, hydroxamic acids are rather nonspecific zinc binders, and many of them hit human enzymes such as HDACs and MMPs. Thus, the researchers turned to fragments to find new metallophilic starting points.
The 1152 members of the Vernalis fragment library were screened against LpxC using three NMR experiments: STD, WaterLOGSY, and CPMG in pools of six. This yielded a remarkable 252 hits in at least one assay. These were retested individually and for competition with a substrate pocket-binding small molecule, resulting in 28 hits, two of which were advanced.
A crystal structure of compound 6 bound to LpxC suggested that adding a hydroxyl group could make additional interactions with the protein, and this was confirmed in the form of compound 10. Further fiddling in this region of the molecule was not successful, and the phenyl ring did not provide good vectors to a hydrophobic tunnel. However, replacing the phenyl with a more shapely piperidine yielded compound 17. Although this molecule had slightly lower affinity, it did provide a better starting point for further optimization, ultimately leading to compound 21, with low nanomolar potency against LpxC. Unfortunately, this and other members of the series showed only weak antibacterial activity.

Compound 9 was weaker than the other fragment starting point, but making and testing related compounds led to improved binders such as compound 27. This was the first molecule in this series to be structurally characterized, and crystallography revealed that the imidazole was making a single interaction with the zinc at the heart of the LpxC active site. Adding a hydroxyl led to bidentate chelator 29 (i.e. two interactions with the zinc) that had better activity, and further structure-based design ultimately led to low nanomolar inhibitors such as compound 43. In contrast to the other series, this one did show antibacterial activity, and the researchers eventually discovered molecules with in vivo efficacy. Both series were also selective against a small panel of human metalloproteases.
This is a nice fragment to lead story (expect it to be included in the next compilation). As the researchers note, it provides two important lessons. First, fragments can provide multiple different starting points for a target. Second, because fragment libraries tend to be small, it can be valuable to take some time to refine a fragment before launching into fragment growing or merging. Indeed, compound 38 (itself fragment-sized) contains only four more atoms than the initial fragment hit, yet has more than a thousand-fold higher affinity. During lead optimization you often need to add molecular weight, lipophilicity, and possibly polar atoms, so it is crucial to get the core binding elements as good as possible.

30 November 2020

Bioisosterism surprises

The concept of bioisosterism is central to medicinal chemistry. Essentially, one functional group is replaced by another which has similar activity but a different chemical structure. This might be done for a variety of reasons: improving pharmaceutical properties, enabling new analogs, or inventing around existing intellectual property. Most medicinal chemists are familiar with common bioisosteres, such as replacing a carboxylic acid with an acyl sulfonamide. But what about replacing a carboxylic acid with an amidine? This and other surprising examples are provided in a new Angew. Chem. paper by Gerhard Klebe and colleagues at Philipps Universität Marburg.
The researchers focused on fragments binding to the hinge region of protein kinase A (PKA), a well-characterized and easily crystallized kinase. As we noted a couple weeks ago, most kinase inhibitors bind to the so-called hinge region, where the adenine ring of ATP normally sits. Protein backbone amides typically make one to three hydrogen bonds with inhibitors. The researchers chose 19 simple fragments, each containing an aromatic ring and various substituents, soaked these into crystals of PKA, and obtained high-resolution (between 1.12 and 1.82 Å) structures. They also experimentally measured the pKa values of each fragment.
All except two of the fragments made one or two hydrogen bonds to a backbone amide NH and/or carbonyl oxygen, but the moieties that did so varied dramatically. Benzamide, with its hydrogen bond accepting carbonyl oxygen and hydrogen bond donating primary amide, is a quintessential hinge-binder, but surprisingly benzoic acid bound in a similar fashion. The measured pKa of this carboxylic acid is 4.01, yet the acid serves as a hydrogen bond donor, suggesting that it is protonated in the active site of the enzyme.
On the other end of the acidity spectrum, a substituted benzamidine fragment with a pKa of 10.78 bound in the neutral form, with a normally charged nitrogen atom serving as a hydrogen bond acceptor. In fact, the binding mode it assumes is identical to that of benzoic acid.
These and several other examples illustrate that protonation states of ligands in active sites can be very different from what one would predict based on calculated or even measured pKa values. There are of course limits: an amidine with a measured pKa of 11.32 avoids the hinge and instead interacts with an aspartic acid side chain.
One quibble is that the researchers did not seem to consider hydrogens on carbon atoms as potential acceptors; these are increasingly recognized as important, including in kinases. One pyridine fragment shown may have a CH in close proximity to a carbonyl, but it is difficult to tell from the figures, and the coordinates have not yet been released.
Another omission is the lack of quantitative information about binding energies. Just because benzoic acid and a benzamidine bind identically does not mean they have the same affinities. That said, Gerhard Klebe warned last year of the dangers of putting too much stock in thermodynamic measurements.
These issues aside, this is a nice analysis and should serve as a useful reminder to medicinal chemists that bioisoteres can be quite unexpected. And once the structures are released in the pdb, they will provide a useful resource for modelers seeking to recapitulate crystallographic data.

23 November 2020

Massive crystallographic drug screen against SARS-CoV-2 main protease

As of November 23, more than 58 million people worldwide have contracted COVID-19, and more than 1.3 million have died. Each of these numbers is roughly two orders of magnitude higher than in this post published exactly eight months ago. Progress towards vaccines and biological treatments has been stunningly fast, but small molecules could still play a role. Towards this end, Sebastian Günther, Alke Meents, and nearly 100 collaborators from the Center for Free-Electron Laser Science at DESY and multiple other institutions have just posted a preprint on bioRxiv.
The researchers were interested in drug repurposing, in which approved or clinical-stage molecules are tested against a new target. Typically this is done in some sort of biochemical or cell-based assay, but in this case the researchers chose crystallographic screening against the main protease (Mpro) from SARS-CoV-2. An independent fragment screen against the same target was published recently in Nat. Comm. (I wrote a companion Comment, and both articles are open-access.)
The current screen of 5953 compounds may be the largest crystallographic screen in history, and the first I know of that used drug-sized molecules rather than fragments. Even more impressive, all compounds were co-crystallized with Mpro, a much more tedious process than the usual soaking. The advantage of co-crystallization is that the protein is more able to change conformation in response to compound binding, but the disadvantage is that small molecules may prevent crystallization. Ultimately 3955 compounds allowed crystal formation, of which 3228 produced crystals that diffracted better than 2.5 Å, and 1196 produced usable datasets. The result? Just 37 unique binders, or 0.6%. Comparing this to the 96 fragment hits from the smaller fragment library is complicated by differences in methodologies, but it does seem likely that molecular complexity played a role in the lower hit rate: the median molecular weight of the drugs screened, 366.5 Da, comfortably exceeds fragment space.
Among the 37 binders, only 29 gave sufficiently well-resolved electron density to determine binding modes. Of these, ten bound covalently. The catalytic cysteine seems particularly reactive, as evidenced by the fact that seven structures showed maleate – a common pharmaceutical counterion – covalently bound. One of the covalent molecules, calpeptin, is a cysteine protease inhibitor that had previously been reported to be active against SARS-CoV-2, but the others are less predictable. In addition to the active site, some molecules bound to two possibly allosteric sites.
Crystallographic hits were tested for inhibition of viral replication in cells. Ten were active, and a few (calpeptin, pelitinib, and isofloxythepin) had single digit micromolar activity. Interestingly, despite being designed as a covalent kinase inhibitor, pelitinib binds noncovalently. In contrast, isofloxythepin, a non-covalent dopamine receptor antagonist, binds covalently.
In addition to the cell-based screen, many of the compounds also showed binding by native electrospray ionization mass spectrometry (ESI-MS). However, as we’ve noted previously, the correlation between affinity and ESI-MS binding can be tenuous. It would be nice to see the affinity or activity of the compounds via a more quantitative method. Indeed, the researchers note that none of the non-peptidic molecules had previously been reported as Mpro inhibitors, so they may be quite weak. Another problem is that – in contrast to the crystallographic fragment screen – none of the coordinates seem to have been released yet. Hopefully this will be rectified when the paper is formally published.
This campaign is yet more evidence that crystallography has come into its own as a primary screening methodology. The researchers note that they “now routinely measure 450 datasets per day,” with a goal of reaching 1000. Whether or not these results impact the course of COVID-19, the techniques developed will likely impact future drug discovery efforts.

16 November 2020

Kinase fragments galore: a free virtual collection

Kinases hold a special place in fragment-based drug discovery. Vemurafenib, the first approved FBDD-derived drug, targets a kinase, as do more than a third of fragment-derived drugs to enter the clinic. These efforts have produced a wealth of knowledge, and in a new paper in J. Chem Inf. Mod. Andrea Volkamer and collaborators at Universitätsmedizin Berlin and Bayer have extracted thousands of virtual fragments and made them freely available in a database called KinFragLib.
The researchers started with a prior database called KLIFS, which compiles thousands of crystal structures of kinases bound to small molecules. Kinase inhibitor binding modes are classified into several  types, and to keep things simple the focus here was on Type I and Type I1/2. Both bind to the active, DFG-in form of the kinase, the only difference being that Type I1/2 binders extend into a back pocket. A total of 2801 crystal structures were selected for analysis.
Next, the ligands were computationally fragmented using a methodology called BRICS (Breaking of Retrosynthetically Interesting Chemical Substructures). Molecules such as ATP and other substrate analogs were discarded to keep the focus on drug-like compounds, and some particularly large, complex molecules such as staurosporine could not be fragmented. The researchers were particularly interested in molecules that bind to the so-called hinge region, where the adenine moiety of ATP binds, so the few ligands that did not bind here were also removed. This reduced the total number of structures to 2553, which yielded 7486 fragments.
The kinase active site was divided into six sub-sites: the adenine pocket, solvent-exposed pocket, front pocket, gate area, and two back pockets. Each of the fragments was then assigned to one sub-pocket. More than 80% of the original (unfragmented) ligands bound to two or three sub-pockets, while another 13% bound to four sub-pockets. Just 5% of the original ligands bound only to the adenine sub-site, but these 127 ligands – with an average of 15 non-hydrogen atoms – could be quite interesting as crystallographically validated fragments.
Various analyses of the fragments binding in each of the sub-pockets reveal trends. Those binding in the adenine sub-pocket tend towards more hydrogen bond donors and acceptors than those in other pockets, as expected. The shapeliness of fragments binding in the various sub-pockets is not quantitatively analyzed, though the interested reader could run these calculations. The 50 most common fragments for each sub-pocket are presented as figures in the supporting information.
Aside from extracting interesting cheminformatic trends, what else can you do with these fragments? The researchers took a subset of 624 rule-of-three compliant fragments and recombined them to generate 6,720,637 distinct molecules. The vast majority of these appear to be novel, and among the 218 that had previously been reported in ChEMBL, more than 20% were potent (IC50 ≤ 500 nM) kinase inhibitors.
With the inexorable increase in docking speeds, this virtual collection of fragments will be useful for building even larger libraries and using them to find ligands for new kinases. And, as the researchers point out, the collection could be useful for fragment growing or merging to new experimentally identified fragments. This is a resource that should be broadly useful for the community.

08 November 2020

From noncovalent fragment to reversible covalent CatS inhibitor

We noted just a couple weeks ago that covalent fragment-based approaches have been on a tear. Much of the recent focus has been on irreversible inhibitors, but as we discussed back in 2013 there is much to be said for reversible covalent molecules too. These are the subject of a new paper in J. Med. Chem. by Markus Schade and colleagues at Grünenthal GmbH.
The researchers were interested in cathepsin S (CatS), one of 11 members of a family of cysteine proteases. The enzyme has been implicated in a laundry list of diseases, from arthritis to neuropathic pain to Sjögren’s Syndrome, and indeed a few inhibitors entered clinical trials in the early twenty-first century. However, selectivity turns out to be essential: inhibiting the related cathepsin K can lead to cardiovascular problems and stroke. Molecules that appear selective often  contain a basic nitrogen and so can accumulate in lysosomes, achieving sufficiently high local concentrations to inhibit CatK.
CatS is a small (24 kDa) enzyme, ideal for protein-observed NMR. An 15N-HSQC screen of 1858 noncovalent fragments yielded 18 hits, all of which showed similar chemical shift perturbations (CSPs) suggesting binding in the S2 pocket. X-ray crystallography was successful for three fragments, confirming that they do indeed bind in the S2 pocket. Appealingly, this region of the protein is structurally different from the other cathepsins, suggesting a route to selectivity.
The sulfonamide moiety of compound 1 (blue) binds in a very similar fashion to the sulfone of a previously reported reversible covalent inhibitor, compound 16 (red). Growing compound 1 to compound 37 led to a significant boost in potency, and crystallography revealed that the binding mode remained the same.

At this point the researchers sought to remove a few heteroatoms as well as introduce the nitrile warhead from compound 16, yielding compound 39b. Surprisingly, this molecule was no more potent than the non-covalent precursor. However, fragment growing into the S3 pocket yielded a massive boost in potency in the form of compound 44. Further SAR and crystallography revealed that much of the increased affinity is due not to specific interactions in the S3 pocket but rather to a hydrogen bond between the newly introduced amide proton and a main chain carbonyl of CatS. Compound 44 is also highly selective against CatK and CatB, showing negligible inhibition of either at 10 µM. Unfortunately, cellular potencies of representative compounds were down by more than three orders of magnitude, likely due to low permeability.
While there is still some way to go to establish whether these molecules will succeed where others have failed, this is nonetheless a nice case of fragment-assisted lead discovery And while one can certainly argue that it would have been possible to derive compound 44 from compound 16 through classical medicinal chemistry, fragments clearly helped.

02 November 2020

Poll results: Conferences in the age of COVID-19

Will you go to a conference in person next year? Our latest poll addressed that question and more. The 6-question survey was conducted on Crowdsignal from September 13 to October 31 and was answered by 121 respondents, 116 of which answered all the questions. My personal thoughts on two of the more significant virtual conferences I’ve attended are here and here.
The first question, answered by everyone, was “Under what conditions would you attend an in-person conference?

Two-thirds of 119 respondents to the question “Have you attended a virtual conference since the beginning of the pandemic?” had done so, and most of them were satisfied.

Most respondents would consider attending a virtual conference, though fewer would consider presenting.
And nearly half of 116 respondents expressed some hesitation to the question, “Are you comfortable having your presentation recorded and viewed later by those who missed the live event?” 

The last question (“Where do you reside?”) revealed a remarkably diverse group from 21 countries.

Overall the main takeaways are that people are comfortable attending and presenting at virtual conferences, but that many are concerned about having presentations recorded and shared. This makes sense: one of the nice aspects of conferences is that you can present preliminary information and discuss emerging (perhaps even half-baked) ideas and ask off-the-wall questions. Still, until our industry comes through with an effective vaccine – or our governments and fellow citizens succeed in reducing transmission – virtual conferences will have to suffice.
Finally, on the subject of polling, the US has a rather important election tomorrow. If you are eligible and have not done so yet, please vote. And to everyone else, please wish us luck.