A very useful list: common linkers and bioisosteric replacements

Last week’s post highlighted an example of fragment linking, which despite being less common than fragment growing can still be effective. But how do you choose the linker? We’ve previously written about the most common rings found in drugs. In a new Bioorg. Med. Chem. paper Peter Ertl and colleagues at Novartis tabulate the most common linkers found in bioactive molecules.

 

The researchers start by defining linkers “as moieties connecting 2 ring systems.” To focus on druglike molecules, linkers could contain no more than eight non-hydrogen atoms total and no more than five consecutive bonds between the two ring systems. This means that para-disubstituted phenyl or 1,4-disubstiuted butyl would both be considered in the analysis, but longer linkers such as this recent example would not.

 

Molecules were extracted from the databases ChEMBL and ZINC, yielding a total of 1686 unique linkers. Various descriptors were calculated for all, which in addition to size and length included the number of heteroatoms and electronic properties. Bioactivity data for molecules in ChEMBL was used to assess which replacements were most frequently tolerated. If one linker could be replaced by another without causing a drop in affinity (or inhibition, etc.), the two linkers were considered to be bioisosteres.

 

So, what are the most common linkers? A single methylene is the most common, followed by an amide bond. I was surprised that, of the 40 most common linkers, only five are rings: para-disubstituted phenyl, 1,4-piperzine, 1,4-piperidine, 1,2,4-oxadiazole, and meta-phenyl, in that order. Not coincidentally, phenyl rings, piperidines, and piperazines are also the most common rings found in drugs, according to an analysis last year.

 

Last year we highlighted a paper from the Ertl group that included a link to a “Ring Replacement Recommender,” which suggests bioisosteric replacements for any ring. Alas, there is no “Linker Replacement Recommender,” but the new paper does provide a “bioisosteric replacement network,” which is a full-page 10 x 15 grid with the 150 most common linkers arranged such that nearby linkers are likely to be bioisosteric. For example, para-phenyl is adjacent to 2,5-thiophene and quite some distance from sulfone. These make sense, but there are also less obvious examples: the table suggests that a 1,4-pyrazole makes a good replacement for a carbamate.

 

The next time you’re doing SAR, it may be worth consulting the bioisosteric replacement network for ideas.

Crystallographic covalent fragment screening – but why?

Once considered a time-intensive method justified only by the most-promising molecules, crystallography is fast becoming a routine technique for general fragment screening. Just last month we highlighted an example where crystallographic hits were successfully used in virtual screens. In a new J. Med. Chem. paper, Jeffrey St. Denis, Benjamin Cons, and colleagues at Astex describe a crystallographic screen with a covalent fragment library.

 

Driven in part by the rapid development of KRAS^G12C inhibitors such as sotorasib, covalent fragments are continuing to increase in popularity. To screen, you first need a library, and here the researchers assembled a set of 114 molecules, most of which were commercially available from Enamine. The majority (101) contained an acrylamide moiety, the most common warhead in approved covalent drugs. The library is rule of three compliant and fairly shapely, as assessed by deviation from planarity. Laudably, the structures of all 114 library members are disclosed in the supporting information.

 

In addition to physicochemical properties, the library was also characterized experimentally. Importantly, it seemed to be quite stable when stored frozen at -80 °C in deuterated DMSO for up to a year. Most (83%) of the compounds were soluble to at least 3 mM in aqueous buffer. And just over half of the molecules withstood a test for reactivity with 5 mM glutathione, the main nucleophile found in cells, assessed as >90% remaining after 30 minutes at 37 °C.

 

As a test case, the researchers screened the kinase ERK2, an oncology target for which Astex previously developed the non-covalent clinical compound ASTX029. Soaking ERK2 crystals in 50 mM of each fragment led to 29 bound ligands, a whopping 25% hit rate. Of these, 16 compounds bound to C166, a cysteine near the hinge region in the active site. Unsurprisingly, less reactive fragments made hydrogen bonding interactions with the protein, while the more reactive fragments did not and also had poor electron density, suggesting non-specific alkylation. As we noted in 2014, distinguishing specific from non-specific binding is a general challenge when screening irreversible fragments.

 

Among fragments making specific interactions, compound 9 was chosen for further study due to its low inherent reactivity and interesting structure. The molecule showed no inhibition of the protein after 1 hour at 1 mM, but scaffold hopping and fragment growing led to compound 22, with low micromolar inhibition after 1 hour. (A back of the envelope calculation suggests k_inact/K_i ~ 25 M^-1s^-1, though I wish the researchers had reported this.)

 

This is an interesting paper, and it is useful to get structural information up-front, but I can’t help thinking that it would be more efficient to use an alternative method for the initial screen. As the researchers acknowledge, many covalent fragment screens use intact-protein mass spectrometry. For example, in a paper we highlighted in 2019, nearly 1000 fragments were screened against ten different proteins. Moreover, as we noted last year, the ability of crystallography to find such very weak hits can be a distraction as much as a blessing.

 

The current paper promises to “report on the further developments and observations pertaining to our electrophilic library in due course.” I look forward to seeing more.

What rings are found in drugs?

Recently we highlighted the “Ring Replacement Recommender,” which provides suggestions for how to improve affinity by replacing one ring with another. The recommendations are based on an analysis of hundreds of thousands of molecules. But what about the rings found in actual drugs? This is the focus of a J. Med. Chem. paper by Richard Taylor and collaborators at UCB and Bohicket Pharma Consulting.

 

The researchers examined FDA-approved and investigational drugs with disclosed structures as of January 2020. These were fragmented into component “ring systems” for analysis. (Ring systems include not just monocycles but fused rings, such as purine. For example, sotorasib consists of four ring systems: benzene, pyridine, piperazine, and pyrido[2,3-d]pyrimidin-2-one.) More than 90% of drugs contain at least one ring.

 

Approved drugs have just 378 unique ring systems in total – a small increase from when the researchers examined approved drugs in 2014. The phenyl ring is found 727 times, with pyridyl (86 examples) a distant second, followed by piperidine (76 examples) piperazine (65 examples) and cyclohexane (47 examples). After that the numbers drop off sharply, with pyrazine in 50^th place with just six examples and fluorene in 100^th place with three examples.

 

Investigational drugs at first appear to be more diverse, with 450 unique ring systems, 280 of which are not found in approved drugs. Of these 280, pyridazine is the most common, with nine examples, followed by oxetane, with seven, but things quickly become less common from there, with 271 of the ring systems found just once. In contrast, ring systems found in drugs are found in multiple compounds, and in fact two thirds of investigational drugs only contain previously used ring systems.

 

Many of the new ring systems are closely related to those found in approved drugs, with nearly half differing by at most two atoms. Perhaps because of this the overall properties of the ring systems are similar between approved and investigational drugs, with no significant differences in heteroatom ratio, percentage of sp³ centers, or number of rings per system.

 

What new opportunities exist? The researchers identified nearly half a million synthetically accessible ring systems and winnowed these down to 3902 ring systems that have similar heteroatom ratios to those found in drugs and differ by at most two atoms. This attempt to explore new chemical space is similar to earlier work from the same group (here) as well as that from others (here, here and here).

 

The researchers also examined growth vectors and combinations of rings, the latter by using graph theory. These analyses suggest that investigational drugs do have greater variety. In other words, even if the component rings are shared with approved drugs, they might be combined in new ways.

 

Whether certain ring systems are more likely to fail in the clinic was intentionally not addressed, due to the difficulty of assessing why the failures occurred. For example, drugs can fail for commercial reasons; a company may choose to drop a drug against a particular target rather than be tenth to market. And even when the failure is due to the science, it might not be an indictment of the drug itself. Verubecestat did lower β-amyloid levels in people as designed, but had no effect on Alzheimer’s disease.

 

This paper is a fun read, and it will likely provide ideas for scaffold hopping and library design. It is also a reminder of how much chemical space remains to be explored.

What to make first? A new “Ring Replacement Recommender” provides suggestions

So you’ve run a fragment screen, gotten some hits, and validated them. What then? Looking for in-house or commercial analogs is always a good idea, but if you’re serious about a project you’ll eventually need to do chemistry, for example replacing one ring with another (say, a pyridyl for a phenyl). The possibilities are almost endless, especially if you don’t know how your fragment binds. In a new Eur. J. Med. Chem. paper, Peter Ertl and colleagues at Novartis describe a “Ring Replacement Recommender” to rapidly improve biological activity.

 

To determine which replacements are likely to improve affinity, the researchers turned to ChEMBL, a database of more than 2 million molecules and associated biological activity extracted from tens of thousands of publications. From these, more than 68,000 chemical series were chosen for analysis. Each series had on average 16 members, and at least three. The biological activity of each member of a series was compared with other members of the same series. (Importantly, the researchers intentionally excluded anti-targets such as hERG and CYPs so the tool wouldn’t inadvertently improve binding to these.) Focusing only on ring replacements that were reported in at least five publications led to a set of 26,762 changes. Changes could be as modest as adding a methyl substituent or more elaborate such as changing a single aromatic ring to a fused aromatic-aliphatic ring system.

 

One would think that most changes would have little effect, as had previously been seen in the case of methyl additions. Indeed about 65% of the replacements caused shifts in potency of 2-fold or less, which is probably within experimental error. However, 2860 replacements of 245 rings improved affinity at least 2-fold (averaging 3.5-fold), with 223 cases yielding greater than ten-fold improvements.

 

Analyzing the data further, the researchers found 80 ring systems that frequently led to improvements in affinity, and they suggest these could be used as “universal” or privileged building blocks. Strikingly, 74 of these are aromatic, confirming work from Cohen we highlighted in 2020 that proteins may favor “flat” rather than shapely molecules.

 

The researchers also extracted 9515 drugs and clinical compounds from ChEMBL and examined the component fragments. Of the 80 ring systems in the universal set, 19 are found in 50 or more drugs, with another 37 found in at least 5 drugs. This set may be a particularly attractive go-to list.

 

Importantly, not only are all the replacements available in the Supporting Information, the researchers have created a handy and free online tool. Just click on a ring of interest and the Ring Replacement Recommender provides suggestions, along with the average fold improvement observed and the number of publications used for the calculation.

 

To see how well it works, I looked at a couple recent examples which entailed ring changes. The indole to indazole replacement used in the TLR7/8 work described last month was not suggested by the Recommender, though in that case the researchers had the benefit of a crystal structure. On the other hand, a cyclobutyl to phenyl substitution for SARS-CoV-2-3CLp was correctly predicted to be beneficial.

 

Of course, as we’ve said repeatedly, affinity is only part of the battle in drug discovery, and the researchers emphasize that their recommendations may not improve physicochemical or pharmacokinetic properties. But for the earliest stage of a program, and especially in the absence of other data, it’s worth giving the Recommender a try.

Automated scaffold hopping for fragments

A post last month covered high-throughput virtual screening, but most practitioners of FBLD still start with some sort of (bio)physical screen. These initial hits can’t be expected to be optimal, since the average fragment library contains a few thousand compounds at most. Indeed, as Xavier Barril and collaborators at Universitat de Barcelona and Oxford University write, “fragment hits should be seen as beacons indicating privileged areas of chemical space to be further explored.” They describe one way to expedite exploration in a recent J. Med. Chem. paper.

 

As we noted here, most good fragments make at least one essential interaction (such as a hydrogen bond) to the protein. The approach starts with a structure of a fragment bound to the target of interest, with that essential interaction identified.

 

Next, a virtual library is searched for similar molecules, with the definition of “similar” being rather loose (>50% Tanimoto similarity). Ideally the library is large enough to produce lots of hits; the researchers used ZINC15, which contains >15 million ostensibly commercial compounds. Also, only molecules within two non-hydrogen atoms of the starting fragment are considered. In other words, a fragment with ten “heavy” atoms would yield molecules with 8-12 non-hydrogen atoms. This search is similar though perhaps more permissive than Astex’s Fragment Network (which we wrote about here).

 

All the molecules are then superposed on the initial fragment structure and only those that maintain the key interaction and binding mode are kept. Aboout 500 molecules are then selected to represent the best and most-diverse hits. These are subjected to dynamic undocking (DUck), which weeds out fragments that have weaker interactions. If desired, each of the remaining hits can be subjected to further cycles.

 

To demonstrate the approach, the researchers turned to bromodomains, a popular target class for FBLD. They started with 1XA, a fragment Teddy highlighted back in 2013 that led to a clinical compound against BRD4. The isoxazole moiety makes a hydrogen bond with the side chain nitrogen of an asparagine that normally binds to acetylated lysine residues. After one cycle, 58 molecules were selected, but unfortunately only five were actually available commercially. Compound 3 had similar affinity and ligand efficiency as 1XA, and this scaffold had not been reported as a bromodomain ligand. A crystal structure of compound 3 bound to the first bromodomain of BRD4 confirmed the predicted binding mode.

Three additional successive iterations were conducted to look for more ligands, but experimental confirmation was challenging as overall only 17 of more than 100 ligands selected for purchase were commercially available. (Compound 23 was chosen for custom synthesis as it was related to a family of high-scoring molecules.) Encouragingly, eight molecules were active in a differential scanning fluorimetry (DSF) assay, a technique that works well for BRD4. Crystal structures of two of these were obtained: compound 9 contains an isoxazole moiety like 1XA (and indeed resembles this fragment) but compound 23 is quite distinct.

Overall this looks like a valuable method for scaffold hopping. Not only might the described approach lead to novel molecules, it could provide new growth vectors that may not be accessible from the original fragment. Before jumping immediately into chemistry with your fragment hits, it may be worth trying something like this.

Efficiently searching for fragments

What do you do when you find a fragment? After checking for artifacts and getting as much structural information as possible, the next step is usually to test analogs for improved potency. But how do you go about that? Richard Hall and his colleagues at Astex provide their approach in a recent paper in J. Med. Chem.

Readily available analogs can come from two sources. Larger organizations generally have massive libraries of compounds, and it’s easy enough to order these for testing. There are also plenty of commercial vendors, enabling SAR by catalog. But how do you sort through the millions of possibilities to find those that are most likely to improve potency?

Sub-structure searches are generally the first approach: look for fragments containing a central core, perhaps differently decorated. A nice example of this is described here, where a search for related pyrimidines led to an increase in potency by replacing one atom. Sometimes more dramatic changes are necessary though. Searching for similar molecules that do not share the same core can be successful, as in this case, but often requires multiple searches. Also, particularly for smaller fragments, “similarity” can encompass significant differences.

The Astex researchers have created a computational tool to streamline this search procedure. It is called the Fragment Network, which is a “graph database,” a type of database in which information is stored as nodes and edges – like the webpages (nodes) and links (edges) used in Google searches. In the Fragment Network, each fragment is computationally dissected into component parts (such as a phenyl ring or a hydroxyl), with edges representing the connections between the parts (such as carbon-carbon bonds). The database contains about 5 million compounds of up to 24 non-hydrogen atoms, and these are further annotated as to whether they are available in-house or from more or less reliable vendors.

A search of the Fragment Network – which takes just a fraction of a second – can be customized depending on the goal. A default search returns compounds that are up to two edges away from the query, which can yield quite a large number of compounds, many of which would not come up in a substructure query, as shown for the simple but useful 4-hydroxybiphenyl.

Plodding through lists of compounds can be tedious, and one nice feature of the Fragment Network is that it groups compounds by type – so for example the ring substitutions are grouped separately from the linker replacements. Compounds are also sorted by commonality of replacement: for example, published data reveals that the most common replacement of a methyl group is a chlorine atom, followed by a methoxy group, with an amine way down the list.

The researchers applied the Fragment Network retrospectively to two previously disclosed programs, campaigns against protein kinase B and HCV NS3. In both cases the program identified most of the changes explored by the medicinal chemists on the project, as well as some that were not tested. Of course, often times the best fragments are not available and need to be synthesized, and the grouping of results returned by the Fragment Network quickly highlights these regions of less-populated chemical space.

Those of you who have seen Astex researchers present at conferences will be familiar with AstexViewer, a powerful open-source molecular visualization program. Hopefully the code for the Fragment Network will also be publically released. If not, it might be worth talking to your computationally gifted colleagues to see if they can create something similar. In the meantime, how many of you are using something similar?

Crystallographic screening of soluble epoxide hydrolase

Last year we highlighted a paper from Yasushi Amano and colleagues at Astellas in which they performed fragment-screening on soluble epoxide hydrolase (sEH), a potential target for inflammation and hypertension. A new paper from the same group in Bioorg. Med. Chem. builds on that work and provides some interesting comparisons.

In the first paper, the researchers performed an enzymatic screen with fragments at high concentrations, resulting in a hit rate of around 7.3%, of which 126 of the 307 hits resulted in crystal structures. However, despite this bounty of hits, only 2 new scaffolds were found that bind to the catalytic triad of the enzyme.

Given the success rate with crystallography, the new paper focused on crystallographic screening as the primary fragment-finding method. The researchers chose 800 fragments from their in-house collection with molecular weights between 151-250 Da. To identify new scaffolds, fragments containing amide or urea moieties – known catalytic-site binders – were excluded. The fragments were then pooled into cocktails of 10 and soaked into crystals of sEH, with each fragment at a final concentration of 1 mM. X-ray diffraction data of the soaked crystals resulted in 8 hits. To ensure that nothing was missed, cocktails of the remaining 9 fragments from pools with a hit were retested, but nothing new came up. Although the researchers do not comment on the lower hit rate compared with the original screen, this could be because they were looking specifically for new scaffolds.

Despite the 1% hit rate, the fragments identified were quite interesting, with IC₅₀ values ranging from 52 to 2200 µM. Most of the fragments formed hydrogen bonds to the catalytic triad, but the details differed from reported inhibitors. For example, several fragments contained secondary amines. Fragment 1 (cyan) in particular was well-positioned to reach into two sub-pockets on either side of the catalytic center, so 14 analogs were chosen for screening, resulting in molecules with significantly increased activity, such as compound 9 (magenta).

The crystal structure of compound 9 bound to sEH reveals that it binds in a similar manner as fragment 1. However, the added hydroxyl group is able to make new interactions that were unavailable to fragment 1, and the larger adamantyl group of compound 9 is able to make more hydrophobic interactions than the smaller phenyl ring.

This is a lovely illustration of the gains in both affinity and ligand efficiency that can be had by scaffold-hopping. It is also a nice example of using fragments to explore new chemical space. Finally, it is laudable that all the structural information is deposited in the protein data bank.

Fragments vs PI3K – AstraZeneca’s turn

Last year we highlighted a paper from AstraZeneca in which they used virtual screening to identify fragments that inhibit the p110β isoform of phosphoinositide 3-kinase (PI3K), a potential anti-cancer and antithrombotic target. In two recent papers in Bioorg. Med. Chem. Lett., Fabrizio Giordanetto and colleagues describe the optimization of one of these fragments to a potent, selective molecule with in vivo efficacy.

The first paper describes the initial fragment-to-lead work. A variety of changes to fragment 1 were explored, including adding a lipophilic phenyl group to increase potency (compound 2). At the same time, modifications were explored in the central pyrimidinone ring. Although compound 3 was less active than compound 2, it also had a considerably lower logD. Many additional changes were explored, and ultimately one of the most potent compounds was compound 16. The isomeric compound 22 was less potent, but had significantly better solublility and stability in a microsomal assay.

The second paper describes subsequent optimization, ultimately yielding compound (S)-21. There’s a lot of good medicinal chemistry that I can’t do justice to here, so definitely check out the two papers themselves. Compound (S)-21 is potent, selective against the related kinase p110α, and shows good activity in a dog model of platelet aggregation without causing an increase in bleeding time.

One of the nice things about this work is the fact that the researchers used a fragment-hopping approach and were not focused on potency to the exclusion of all other properties. Although one could argue that this is simply good medicinal chemistry, it can sometimes fall into the category of what Mike Hann has memorably christened “unknown knowns,” a trap this team avoided.

Hopping to selective nNOS inhibitors

Richard Silverman, the discoverer of pregabalin, has published a trifecta of papers describing the use of his “fragment hopping”strategy to discover selective, potent, and pharmacologically active inhibitors of neuronal nitric oxide synthase (nNOS).

nNOS is itself part of a triad of closely related enzymes: nNOS, iNOS, and eNOS. All synthesize the radical signaling molecule NO from arginine, but their tissue localizations and functions differ considerably. Both nNOS and iNOS have been pursued as drug targets, with nNOS implicated in a variety of diseases of the central nervous system. However, as eNOS is involved in maintaining blood pressure, identifying inhibitors selective for the desired NOS is essential. Doing so is difficult because all three enzymes share very similar active sites: of the 18 amino acids close to the substrate binding site, 16 are identical, and one of the two residues that vary has its side chain facing away from the substrate binding site.

The Silverman group has been working on selective nNOS inhibitors for over a decade, and succeeded in generating remarkably selective inhibitors, such as the compound shown on the left. But with its peptidic nature and dramatic CLogP, this compound is a long way from a drug. Enter “fragment hopping.” First, a known inhibitor is computationally deconstructed into its minimal pharmacophores; these can be as rudimentary as a positive charge or a hydrophobic spot. Next these virtual fragments are rebuilt into new virtual molecules; further computation weeds out molecules that are likely to be metabolically unstable or toxic. Finally, the best molecules are synthesized and tested. The process is computationally intensive, and I’m simplifying it greatly here, but the results, first published in J. Am. Chem. Soc. last year, are impressive: the dipeptide was transformed into a smaller, more drug-like molecule (center in figure), with increased ligand efficiency. It also retained selectivity over iNOS and eNOS.



In the latest issue of J. Med. Chem., Silverman and colleagues report using the technique again, this time to not only improve potency, but to improve the drug-like properties as well. The resulting molecule (right in figure) still maintains good selectivity, but now also boasts a respectable CLogP. Consistent with this, the molecule shows efficacy in a rabbit model of cerebral palsy. For a great review summarizing the long march to these compounds, check out this just released Acc. Chem. Res. paper.

What I really like about this work is that the researchers use fragment-based methods to actually improve the pharmaceutical properties of their molecules. Much of the focus of FBDD has been in discovery of early leads; these papers show that the approach can be useful downstream as well.

And finally, we can’t let today pass without sending out a Happy 200th Birthday to Charles Darwin (as well as Abraham Lincoln). As Teddy noted in an earlier post, “it’s evolution baby.”