Flatland: still a nice place to be

In 2009 we highlighted a paper reporting that approved drugs have a higher fraction of sp³-hybridized carbon atoms than discovery-phase compounds. Perhaps focusing on molecules with a high Fsp³, the ratio of sp³-hybridzed carbons to total carbons, would lead to greater success. Or not: a new analysis in Nat. Rev. Chem. by Ian Churcher (Janus Drug Discovery), Stuart Newbold, and Christopher Murray (both at Astex) finds that the relationship has not held up.

 

The 2009 “Escape from Flatland” paper was widely discussed at conferences and has been cited more than 3000 times. But according to the authors of the new study, most of these citations are from papers describing new synthetic methodologies rather than from papers discussing medicinal chemistry.

 

And not all the attention has been positive. As we noted in 2013, Pete Kenny and Carlos Montanari reanalyzed the data and found that an apparent correlation between Fsp³ and solubility disappeared when plotting all discrete data points instead of binned data.

 

More recently, we highlighted a paper that found no significant difference between the shapeliness of drugs, as assessed by their principal moment of inertia (PMI), and the shapeliness of small molecules in the ZINC database.

 

The new paper looks at Fsp³ values for drugs approved during various time periods. Among 980 drugs approved up to 2009, the average Fsp³ was 0.458. However, of the 431 drugs approved after 2009, the average Fsp³ has dropped to 0.392. The researchers speculate that this (statistically significant) decrease may be due to an increase in the number of kinase inhibitors, which are usually highly aromatic, as well as an increase in the use of metal-catalyzed cross coupling reactions.

 

In my analysis of the 2009 paper, I asked whether higher Fsp³ ratios would lead to lower hit rates, and indeed this seems to be the case, as shown in a paper we discussed in 2020. Thus, if you pursue difficult targets, you may increase your chances of finding hits by screening molecules with lower Fsp³ ratios. Also, multiple studies, including one published just last month, have found no correlation between the shapeliness of a fragment (as defined by deviation from planarity, or DFP) and the shapeliness of the resulting lead, so there appears to be no penalty to starting with a flattish fragment. 

 

The researchers conclude that their “analysis of drug development trends over the last 15 years suggests that Fsp³ may not have been a useful metric to optimize.” Importantly, the supplementary information includes a list of >1400 approved drugs and >1500 investigational drugs along with associated properties, so you can do your own analyses.

 

In the end, generalizations will only get you so far, and may even lead you astray. At least for now, there are few shortcuts in the long slog of experimental studies necessary to discover a drug.

Stitching together fragments with Fragmenstein

As we noted just last week, crystallography has unleashed a torrent of protein-ligand complexes, especially fragments. Historically a single structure might be used for fragment growing, but so many structures present an embarrassment of riches, with sometimes dozens of fragments that bind in the same region. Merging or linking these fragments can be done manually, as seen here and here, but how to do so when the binding modes are partially overlapping is not always intuitive. In a new open-access J. Cheminform. paper, Matteo Ferla and colleagues at University of Oxford and elsewhere describe an open-source solution called Fragmenstein.

 

We briefly described Fragmenstein in 2023, where it was used to combine pairs of low-affinity fragments bound to the Nsp3 macrodomain of SARS-CoV-2 to generate sub-micromolar inhibitors. The current paper describes the platform in detail.

 

Fragmenstein starts by taking two (or more) structures of fragments bound to a protein and virtually combining them. This is done by collapsing rings to their individual centroids, stitching these together along with their substituents, and then re-expanding the ring(s) so the substituents will be close to where they were in the initial fragments. This process produces a surprising array of molecules beyond the obvious. For example, if one fragment contains a phenyl ring and the other fragment contains a furan ring, the stitched molecule might just contain the phenyl (if the two rings bind in nearly the same position), a benzofuran (merging the rings), a phenyl ring linked to a furan by one or more atoms, or even a spiro compound if the rings are perpendicular to one another.

 

In silico approaches sometimes suggest molecules that are synthetically challenging to make, but Fragmenstein can also be used to find purchasable analogs.

 

Next, the new molecules are energetically minimized, first by themselves and then while docked into the protein. In contrast to other docking programs, which might allow molecules to sample thousands of different conformations and sites in a protein, Fragmenstein maintains the new molecule in a similar position and orientation to the initial fragments, with the assumption that these have already identified energetically favorable interactions.

 

The researchers successfully applied Fragmenstein retrospectively to several targets. The COVID Moonshot (which we discussed here) crowd-sourced molecule ideas for the SARS-CoV-2 main protease based on structures of bound fragments. Of 87 ligands that had been crystallographically characterized and were designed based on two fragments, Fragmenstein successfully (RMSD < 2 Å) predicted the binding mode for 69%.

 

Fragmenstein can even be used for covalent ligands, as shown for the target NUDT7, which we wrote about here. Merging two fragments led to compound NUDT7-COV-1, and the RMSD between the Fragmenstein model and the crystal structure was an impressive 0.28 Å.

 

Of course, as the researchers acknowledge, the number of possible analogs might be daunting, and deciding which to make or buy is not necessarily straightforward. Also, Fragmenstein assumes that the fragments themselves are making productive interactions with the protein, which may not be the case, as we suggested here. Still, the tool is open-source and worth trying, especially if you are swimming in crystal structures.

The thousandth Practical Fragments post!

As the title states, this is the 1000^th post at Practical Fragments. This blog was conceived in 2008 over drinks at the Third Annual CHI Fragment-based Drug Discovery Conference. (Don’t miss the twentieth in April!) Teddy Zartler said he was planning on starting a blog and asked if I wanted to join. In July of 2008 Teddy wrote the first post, and every month since then has seen at least a couple new ones. I thought it would be fun to look back briefly on the past 16+ years.

 

Methods

The very first Practical Fragments post asked what screening methods people use, and this eventually led to five polls on the topic, the latest of which published just a few months ago. In our first formal poll, in 2011, the average respondent used 2.4 techniques. Today that number has grown to 5 due to the increased recognition that different methods have different strengths and weaknesses.

 

By far the biggest winner among methods has been X-ray crystallography; it jumped from sixth place in 2011 to first place in both 2019 and 2024. Crystal structures have long been prized in drug discovery, but the dramatic increases in throughput and automation over the last decade mean more structures are more available to more users.

 

Computational methods too have improved spectacularly. In 2009 we highlighted an in silico screen of around 67,000 fragments which yielded ten micromolar inhibitors. Today, screening multibillion compound libraries is becoming routine, and artificial intelligence is likely to enable even more opportunities.

 

Pitfalls

One of the reasons that it took so long for FBLD to develop was the myriad artifacts that can haunt screens run at high concentrations. For example, compound aggregation was not widely recognized until the first decade of this century, and even today too many papers are published without checking for this pathological phenomenon.

 

Similarly, pan-assay interference compounds, or PAINS, were not defined until 2010. Scientists at large companies have long known to steer clear of certain chemotypes. Now academics and folks in startups are more aware of problematic substructures, even if Dr. Saysno objects.

 

Long-time readers may recall a series of posts on “PAINS-shaming,” where we highlighted (lowlighted?) papers that lacked appropriate selectivity or mechanistic studies. Occasionally this led to productive discussions, as in this example where an author and journal editor contributed to the comments. But with the increasing use of metrics measuring social media engagement to rank articles I’ve decided that blogging about them may inadvertently reward shoddy science. If you’re looking for most of the things that can go wrong in a screen, check out this open-access review by Ben Davis and me.

 

Covalent craze

One prominent mechanism of PAINS is indiscriminate covalent modification of proteins. For many years drug hunters actively avoided covalent modifiers for fear of off-target modifications and their potentially toxic effects. Indeed, the first several mentions of covalent compounds at Practical Fragments were in the “things to avoid” category. We discussed reversible covalent modifiers in 2012 and 2013, but it wasn’t until 2014 that we wrote about intentionally irreversible fragments.

 

How times have changed! The success and safety of targeted covalent kinase inhibitors has fueled enthusiasm for covalent drugs in general, creating opportunities for fragment-based approaches. Indeed, as we discussed here, both reversible and irreversible fragment-based screens were used in the discovery of the first approved drug targeting the previously intractable target KRAS, and these learnings have been applied at multiple companies to produce an impressive armamentarium against what Darryl McConnell has called “the beating heart of cancer.”

 

To find KRAS inhibitors, researchers screened pure proteins against libraries of covalent fragments. One of the most exciting recent developments in chemoproteomics has been screening covalent fragment libraries in intact cells or cell lysates to find hits against thousands of proteins in their native environment. We first wrote about this approach in 2016, and last year we highlighted the first drug to enter the clinic from covalent screening in cells.

 

And all this is just the beginning: each of our past four annual “review of reviews” posts has featured between three and six papers focused on covalent fragment-based drug discovery.

 

Clinical compounds

My first blog post in 2008 was a brief mention of a C&EN story on FBLD, in which I noted that “an FBLD drug that reaches the market by 2011 would be a ‘psychological’ victory for the whole FBLD community.” Although I claim no prescience, I was happy to see vemurafenib approved in August of 2011.

 

Indeed, I would argue that FBLD-derived drugs are the most meaningful output and validation of the field. Our first systematic tabulation in 2009 counted just 17 that had entered clinical trials, and today there are more than 60. Like investigational drugs in general, the majority of these have stumbled, but at least eight have been approved by the US FDA, and more are working their way through clinical trials. While eight might seem like a modest number, the number of patients they’ve helped is orders of magnitude greater.

 

Closing thoughts

There are far more themes in a thousand posts than I could summarize in a single one: metrics, induced proximity, and newer methods such as cryo-EM all come to mind. But as this post has already surpassed 1000 words, I’ll wrap it up.

 

One minor frustration has been the sparsity of comments; it sometimes feels as if I’m blogging into the void. That said, I’m pleased that some posts may have led to new research, such as this. And blogging can be its own reward: I sometimes find myself using the “Search This Blog” function on the top right-side of the page when I’m trying to remember a paper from years ago.

 

Since Teddy left the FBLD field several years ago I’ve been writing most of the content, with occasional guest posts (such as this from Glyn Williams). At the current rate it might take a couple decades for Practical Fragments to reach 2000 posts, if we even get there. But for now, I’d like to thank each of you for reading. I hope you enjoy it and that is has, at least occasionally, made your scientific pursuits more practical.

Fragments vs PRC1: toward a chemical probe

E3 ligase proteins conjugate ubiquitin to other proteins, changing their function or targeting them for degradation. Hijacking E3 ligases using PROTACs or molecular glues has become a popular approach for targeted protein degradation, but some E3 ligases can be drug targets themselves. For example, Polycomb Repressive Complex 1 (PRC1) ubiquitylates histone H2A in a process essential for proliferation of acute myeloid leukemia and some other cancer cells. In a new J. Med. Chem. paper, Jolanta Grembecka, Tomasz Cierpicki, and colleagues at University of Michigan Ann Arbor describe their discovery of small molecule inhibitors.

 

As the name implies, PRC1 is a protein complex. The core includes either RING1A or RING1B and one of six other proteins. Thus, the researchers sought to find inhibitors of both RING1A and RING1B. They started with a ¹H-¹⁵N HSQC NMR screen of 1000 fragments in pools of 20, with each compound at 0.25 mM. (This and some other work was described in an earlier Nat. Chem. Biol. paper by the same authors.) Compound RB-1 was found to bind very weakly to RING1B, and chemical shift mapping revealed that it binds in a region important for E3 ligase activity.

 

Scaffold hopping led to the pyrrole-containing compound 1b, which was slightly more potent as well as more soluble than RB-1. Fragment growing on both rings led to compound 1f, and further optimization yielded low micromolar compound RB-2. Crystallography with this compound revealed conformational changes that opened a hydrophobic pocket in the protein. Although no electron density for RB-2 was observed, the researchers used the crystal structure in combination with NMR experiments to develop a binding model. This facilitated further optimization, ultimately yielding RB-4, the most potent compound reported, with low micromolar affinity as assessed by isothermal titration calorimetry.

As noted above, it is important to block both RING1B as well as RING1A, and the researchers tested their compounds against both proteins using an AlphaScreen competition assay. Most compounds were equipotent against both proteins, though some showed around two-fold greater affinity for RING1A.

 

Subsequent experiments demonstrated that some of the compounds could block ubiquitylation of H2A in an in vitro assay. Importantly, compound RB-4 functionally inhibited three different PRC1 complexes having either RING1A or RING1B along with one of two other protein partners. Treatment of leukemia cells with the compound also led to changes in gene expression and lower levels of ubiquitylated H2A consistent with PRC1 inhibition.

 

This is a nice fragment-to-lead story, particularly given the difficult nature of the protein and the absence of co-crystal structures. While compound RB-4 is insufficiently potent to be called a chemical probe, it is nonetheless a well-characterized starting point for further optimization.

Berotralstat: an overlooked fragment-derived drug

At the end of 2023 I mentioned that a paper by Dean Brown listed berotralstat as a fragment-derived drug. Readers will notice this molecule does not appear on our “fragments in the clinic” list. Did we miss it? After reading a (2021!) J. Med. Chem. paper by Pravin Kotian and colleagues at BioCryst, I believe the answer is yes.

 

Hereditary angioedema (HAE) is a rare genetic disease caused primarily by deficiencies in a protein that inhibits a serine protease called plasma kallikrein, or PKal. Drugs had already been developed to replace the inhibitor protein, but these need to be injected or infused. Since PKal is an enzyme, the researchers sought to make a small molecule inhibitor that could be taken as a pill.

 

BioCryst had developed an earlier drug called BCX4161, which is potent but has poor oral bioavailability. To find a better molecule, the researchers turned to the rich literature around serine protease inhibitors, which led them to make compound 2, a fragment of previously reported inhibitors of other serine proteases. The protonated benzylamine was expected to bind in the S1 pocket of the enzyme, and indeed the molecule did show weak but measurable activity.

 

Fragment growing led to compound 4, with double-digit micromolar activity. Building off the new phenyl ring led to more potent molecules such as compound 13, with low micromolar activity. Further structure-based design eventually led to BCX7353, or berotralstat. The paper provides good descriptions of the design rationale. For example, the fluorine was added to improve permeability, and the nitrile was added to improve the ADME profile. Modeling was used both to improve potency as well as to gain selectivity over other serine proteases. This proved to be successful: berotralstat is a subnanomolar inhibitor of PKal and at least several thousand-fold selective over trypsin and other serine proteases such as thrombin and FXa.

 

The pharmacokinetic properties of berotralstat in rats and monkeys were also good, and according to clinicaltrials.gov the molecule first entered the clinic in 2015. In December of 2020 the FDA approved berotralstat for prophylactic treatment of HAE attacks.

 

This is a nice story, and I agree with Dean that the discovery of berotralstat was “based on a legacy clinical candidate and fragment approaches.” The earlier molecule BCX4161 contained a benzamidine moiety, which was in part responsible for the poor oral bioavailability. Replacing this with a benzylamine fragment from the literature is a classic fragment strategy, and compound 2 is fully compliant with the rule of three.

 

So how was it missed? The abstract only states that berotralstat was discovered “using a structure-guided drug design strategy.” Indeed, the word “fragment” appears precisely once in the paper, albeit in a very telling sentence: “We evaluated these fragments in our PKal^pur inhibitor assay…”

 

From a timeline perspective, the approval of berotralstat makes it the fifth approved fragment-derived drug, after pexidartinib and before sotorasib. I’ll include it in the next update of clinical compounds, along with my standard disclosure that “the list is almost certainly incomplete.” What else are we missing?

Fragment events in 2025

After a bumper year for conferences in 2024, this year is also shaping up to be eventful. 

April 14-17: CHI’s Twentieth Annual Fragment-Based Drug Discovery, the longest-running annual fragment event, returns as always to San Diego. This is part of the larger Drug Discovery Chemistry meeting. You can read impressions of the 2024 meeting here, the 2023 meeting here, the 2022 event here, the 2021 virtual meeting here, the 2020 virtual meeting here, the 2019 meeting here, the 2018 meeting here, the 2017 meeting here, the 2016 meeting here; the 2015 meeting here, here, 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. 

 

May 5-6: Returning after a five year hiatus, Industrial Biostructures of America will be held in Cambridge, MA and includes a session on FBLD. 

June 2-4:  The Eleventh NovAliX Conference returns to the stunning city of Strasbourg. You can read my impressions of the 2018 Boston event here, the 2017 Strasbourg event here, and Teddy's impressions of the 2013 event here, here, and here.

September 21-24: FBLD 2025 will be held in the original Cambridge (UK),  where it was supposed to be held in 2020. This will mark the ninth in an illustrious series of conferences organized by scientists for scientists. You can read impressions of FBLD 2024, FBLD 2018, FBLD 2016, FBLD 2014, FBLD 2012, FBLD 2010, and FBLD 2009. 

 

September 22-25: You'll need to make a tough choice: FBLD 2025 or CHI’s Twenty-Third Annual Discovery on Target. 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 2024 meeting here, the 2023 meeting here, the 2022 meeting here, the 2021 event here, the 2020 virtual event here, the 2019 event here, and the 2018 event here.

 

Finally, from November 11-13 CHI holds its second Drug Discovery Chemistry Europe in beautiful Barcelona. This will include likely tracks on lead generation, protein-protein interactions, degraders, machine learning, and probably several fragment talks. 

   

Know of anything else? Please leave a comment or drop me a note.