26 September 2022

FBLD meets DEL part two: let there be light

DNA-encoded libraries (DEL) are collections of peptides or small molecules attached to DNA tags. In a typical application, libraries are mixed with a protein of interest, non-binders are washed away and those that remain are identified by using PCR to amplify the DNA tags. Two years ago we highlighted an article in which previously identified fragments were merged with molecules identified from DEL. However, because fragments typically have low affinities, screening fragments directly by DEL would seem to be difficult. In a new open-access RSC Medicinal Chemistry paper, Rod Hubbard and collaborators at Vernalis and HitGen describe how to do so. (Rod presented some of this work in April at the CHI DDC meeting.)
To identify weak binders, the researchers turned to photoactivatable fragments that – in the presence of UV light – would bind irreversibly to a nearby protein. Specifically, they used the diazirine tag, which has proven useful in both cell-based screening as well as screens of isolated proteins. Here, the researchers generated two libraries of fragments bound to DNA, with each library member also containing a diazirine tag. The libraries were built using different chemistries and consisted of 15,804 and 23,905 members, small by DEL standards (which often range in the millions) but large by fragment standards.
The PAC-FragmentDEL libraries were incubated against two proteins: the kinase PAK4 and the bacterial enzyme 2-epimerase. Each protein was incubated with both libraries for one hour at room temperature and then treated with ultraviolet light for 10 minutes on ice. Next, the proteins were captured on an affinity resin and washed extensively under denaturing conditions to remove any non-covalently bound library members. Finally, the DNA was amplified by PCR and quantified; any library members that bind to the protein stand out over background.
Of course, there is plenty of opportunity for non-specific binding, so the researchers incorporated several controls, such as omitting the UV-crosslinking step or protein. Moreover, they repeated the experiment in the presence of known high-affinity binders and looked for fragments that were competed.
In the case of PAK4, the researchers identified 301 fragments that could be competed. Eleven of these were further examined (without the DNA tags), all of which demonstrated binding by ligand-observed NMR, and ten of them yielded crystal structures bound to the protein. The examples shown in the paper occupy the hinge-binding site, which the researchers acknowledge is a low bar for fragment screens.
The second target, 2-epimerase, has a more challenging active site, and indeed the hit rate was lower: just 21 competitive fragments were found. But all 9 of those selected for further testing confirmed by ligand-observed NMR, and 5 of them yielded crystal structures.
This paper demonstrates that DEL can be used to identify fragment hits with a fairly low false-positive rate. But do we need yet another fragment-finding method? The researchers point out that PAC-FragmentDEL is fast, with screening and sequencing analysis taking just a few weeks. This also means that fragment libraries can be much larger than for most techniques. Protein requirements are also modest, at around 250 pmol (12.5 mg for a 50 kD protein). They also note that – because of the DNA tag – less intrinsically soluble fragments can be screened, increasing chemical diversity, though one might counter that this could lead to problems down the road.
On the downside, it is not clear whether affinity information can be obtained from the primary screen. Also, the need for a competitive tool molecule could limit choice of targets, as some of the most interesting targets lack any chemical probes. Still, as the researchers note, the competitor could be a peptide or protein, and in a pinch the site of interest could be mutated.
In summary, this looks to be an interesting approach, and I look forward to seeing more applications.

19 September 2022

Crystallography first, then virtual screening: application to PKA

Fragment-based screening is often funnel-shaped: a virtual screen might identify dozens or hundreds of potential hits that are tested in various assays, eventually leading to a few chosen for crystallography. But a paper we highlighted back in 2016 argued that many assays miss genuine hits, and crystallography should be moved to the front of the line. A paper just published in J. Med. Chem. by Serghei Glinca and collaborators at CrystalsFirst, BioSolveIT, Enamine, and elsewhere provides a proof of concept.
The researchers started with a set of 19 crystal structures of fragments bound to Protein Kinase A (PKA) from a campaign we wrote about in 2020. Four diverse fragments were chosen for further study. Importantly, the affinities of these fragments had not been measured; selection was based on the diversity of chemical structures and binding modes.
Next, the crystal structures of the four fragments were used as starting points for four virtual screens using 208,293 Enamine REAL Space fragments (see here for more on these). These were docked using BioSolveIT’s FlexX algorithm, and 50 from each of the four screens were then computationally grown. Just over half a million of these elaborated molecules were then docked, and after clustering, triaging, and visual selection, 106 were chosen for synthesis, of which 93 were delivered and 75 were soluble at 200 mM in DMSO.
The soluble fragments were tested in a functional assay, and 30 of these showed inhibition. Most were weak (double digit micromolar or higher) but fragment EN093 (derived from Frag2) was a low micromolar inhibitor. All of the initial fragments were very weak inhibitors, with at best millimolar activity.

The 75 soluble compounds were also tested in a thermal shift assay (each at 2.5 mM), revealing 29 hits, of which 19 were also active in the functional assay. These included EN093. Interestingly, only one of the initial fragments (not Frag2) showed any activity in the thermal shift assay.
To assess how well the docking performed, 13 of the most active compounds were tested in co-crystallization experiments, yielding 6 high-quality bound structures. These confirmed the virtual screens, with the rmsd for EN093 being 0.74 Å.
Impressively, the whole study, including compound synthesis and crystallographic validation, took just 9 weeks.
This “Crystal Structure First” is conceptually similar to the V-SYNTHES approach we discussed earlier this year, with the difference being that while V-SYNTHES is entirely virtual, Crystal Structure First starts with an actual structure. As the researchers state, “using crystallographically validated fragments and bound ligands for template-based docking can be thought of as introducing a ‘magnet’ to help find the needle in an ever-growing haystack in a more targeted way.”
This is a nice case study, and intuitively it makes sense to start with an experimentally determined structure. Indeed, the increasing number of publicly available fragment structures should be a boon for this approach. That said, it is interesting that most of the molecules made and tested are quite weak, and only two have ligand efficiencies equal to or greater than 0.3 kcal/mol per heavy atom. As we suggested earlier this year, crystallography may find ligands that are just too weak to be useful. Perhaps adding a functional screen before computational elaboration could lead to even more and better binders.

12 September 2022

Growing fragments in silico with FastGrow

Growing fragments is probably the most common approach to improving affinity, and it is immeasurably faster to do this virtually than experimentally. But as anyone who has ever tried can attest, this is often easier said than done. In a new open-access J. Comput. Aided Mol. Des. paper, Matthias Rarey and collaborators at Universität Hamburg, Servier, and BioSolveIT describe a free tool to help.
The application is called FastGrow, and it can be accessed through this web server or the SeeSAR 3D software package. It relies on the “Ray Volume Matrix (RVM) shape descriptor,” which simplifies chemical fragments and protein binding pockets into three-dimensional shapes. This allows extremely rapid assessments of whether a given fragment can fit into a binding pocket. A scoring function called JAMDA assesses interactions beyond simple shapes, such as hydrogen bonds and hydrophobic contacts, and also allows fragments to shift slightly to optimize complementarity with the protein.
One nice feature of FastGrow is that users can input fragments into multiple binding sites with different amino acid conformations, allowing for protein flexibility. You can also specify an important interaction, such as a critical hydrogen-bond, that you prefer to maintain.
To validate the approach, the researchers turned to the database PDBbind and looked for examples in which two ligands with identical cores but different substituents bound to the same protein. They chopped off the substituents from the first ligand and used the resulting fragment as a starting point to try to grow the second ligand. Running 425 of these took just 3 and a half hours and successfully recapitulated the binding mode 71% of the time. This was higher than the popular program DOCK (version 6.9), which seemed to be a pleasant surprise. They attribute the difference to a higher clash tolerance for FastGrow in the initial stages.
For additional validation, the researchers turned to real-world examples of fragment-growing for the kinases DYRK1A/B, which we highlighted last year (here and here). Here too FastGrow outperformed DOCK and was also about five-fold faster when using JAMDA (and 600-times faster without JAMDA, though at some cost in performance).
FastGrow looks to be a valuable tool, and indeed the researchers note that it is currently in use at Servier. There is a lot more detail in the paper and supplementary materials, including the full code for the FastGrow web server and all the underlying data. It would be interesting to compare its performance to the V-SYNTHES approach we highlighted earlier this year.
If you have experience using FastGrow, please leave a comment!

05 September 2022

Is phenotypic fragment screening worthwhile?

Fragment-based drug discovery is almost always target-based. Indeed, not until the development of powerful biophysical techniques such as protein-labeled NMR did FBLD really began in earnest. Phenotypic fragment screens against cells, tissues, or animals are uncommon. In an open-access Front. Pharmacol. paper, Chris Lipinski and Andrew Reaume (Melior Discovery) argue that they should be used more often.
The researchers analyzed all 184,139,678 compounds in the CAS registry with molecular weights between 100 and 999 Da. These were divided into 18 bins (100-149 Da, 150-199 Da, etc.) Next, they calculated the percentage of molecules within each bin with any biological data as evidenced by the “biological study” tag in SciFinder-n.
In terms of raw numbers, fragments are well-represented, with the 250-299 Da bin containing close to 40 million molecules. However, only about 4% of these had any biological data. Molecules with molecular weights between 300 and 549 were abundant and also had considerably more biological data – up to roughly 50% of compounds in the 500-549 Da bin. In other words, people don’t seem to be screening lower molecular weight compounds in biological assays as often as they are screening larger molecules.
The assumption may be that small fragments are not biologically active, but the researchers revisit a classic In the Pipeline post in which Derek Lowe lists 56 drugs with molecular weights equal to or less than that of aspirin (180 Da). Most of these are old drugs, with all but three first reported in the chemical literature before 1980.
The researchers suggest that more effort should go into exploring the biology of smaller molecules, particularly those for which some activity is already reported. They also draw an interesting distinction between two uses of the word pleiotropic. People often say that a drug has pleiotropic effects if it acts on multiple targets; a classic example is imatinib, which hits several kinases in addition to the target BCR-ABL. However, the term pleiotropic originates in genetics and initially referred to one gene having multiple effects. Thus, a drug that acts on a single protein can have multiple effects, as in the case of the PDE5 inhibitor sildenafil.
As an example of a pleiotropic fragment, the researchers discuss MLR-1023, a fragment-sized molecule first discovered in a phenotypic screen at Pfizer in the 1970s. The molecule has shown promise in disease models ranging from atherosclerosis to myeloproliferative neoplasms and was taken into the clinic by Melior in 2014 as an anti-diabetic agent. All of these varied effects seem to stem from the ability of the compound to act as an activator of Lyn kinase. With just 15 non-hydrogen atoms and a molecular weight of 202 Da MLR-1023 is comfortably within rule of three space. Despite its small size, the molecule is a potent activator of Lyn, with an EC50 around 50 nM, giving it a ligand efficiency of 0.66 kcal/mol per heavy atom.
Is MLR-1023 an outlier or an example of an underexplored pool of pharmacological riches? My suspicion is the former. It is rare to find fragments with EC50s < 1 µM, let alone < 100 nM. Moreover, I suspect that many proteins are so difficult to drug that a molecule will need to be well beyond fragment-space – and even rule-of-five space – to have an effect. The protein-protein interaction targeted by venetoclax (MW = 868 Da) immediately comes to mind.
That said, the idea that a large group of tiny molecules is underexploited is worth exploring. For some types of drugs perhaps we don’t need extreme potency: Mike Hann noted a decade ago that the EC50 values of approved drugs average 20-200 nM and cautioned against an “addiction to potency.” And because fragments are likely to have low affinities towards most proteins, they may even be more specific than larger drugs. It will be fun to discover how much room there really is at the bottom.